ValuationMCP
Start free

Release notes

What shipped, by version

Every ValuationMCP release in detail, newest first. Each version keeps its own changelog: the recent releases are listed on top of the full v1.0 launch notes.

v1.0.36

ValuationMCP has introduced a rebuilt Network Overview and a Valuation benchmarks page under InsightsNovember 13, 2026

ValuationMCP has introduced a rebuilt Network Overview and a Valuation benchmarks page under Insights. Overview now reads one row per section, and the valuation distributions moved to their own page. ValuationMCP has also introduced Findings on the Network knowledge graph.

Overview

What needs attention, section by section.

  • What needs attention keeps its five queues and adds a row from your desks and governance: member requests awaiting a response, recruits who shared figures, likely sellers with no conversation, close-and-unprepared owners with no open review, and consent withdrawals in the last 30 days.
  • New rows: Network pulse, Desks (one card per desk you can open), Reports and governance (adoption this month, consent mix, recent activity) and a Valuation snapshot of four medians. Every card links to the page that resolves it and keeps the selected unit.

Insights → Valuations

The benchmarks, on one basis.

  • Valuation benchmarks holds the distributions, opportunity totals, method comparison and method ranges, plus the average quality multiplier and data provenance, all from each practice's latest saved valuation under the same filters.
  • The duplicate all-runs averages block is retired. Groups under five practices still show a dash. Saved outputs only; nothing is recomputed.

Knowledge graph → Findings

ValuationMCP has introduced Findings on the Network knowledge graph.

  • A Findings panel above Layers lists up to six things worth noticing: compound risk, stranded sellers, buyer concentration, data dependency, the most common move, version drift and disconnected firms. Each is a fixed rule over the graph the page already loaded; no sentence is written by a model and no figure is added.
  • Choosing a finding lights its firms and links, dims the rest, frames them and shows a caption with the rule and a link to the page that works on it. It can be shared with ?finding=, and the List view filters to the same set. Groups under five firms show no count, and outputs-only firms are never named.

v1.0.35

ValuationMCP has introduced Network Desks: Practice management, Recruiting, M&A and SuccessionNovember 12, 2026

Network accounts get a new Desks menu with four desks: Practice management, Recruiting, M&A and Succession. Insights shows what's happening; Desks are where you act. Deal Room is now the Diligence Room, with three room types, and member firms get a continuity and succession move card. A firm's own assistant can work incoming recruiting and practice requests step by step, and the public pages explain it all. Every send waits for a person. Engine 1.8.0 unchanged.

Desks

One menu for where you act.

  • The Network menu now has Desks: Practice management, Recruiting, M&A and Succession. Insights shows what's happening; Desks are where you act. Old addresses still work.

Practice management desk

Helping member firms move their numbers.

  • Request figures from a member, suggest moves from its own saved valuation, or offer a named professional. The member chooses Add to plan, Assign or Not now.
  • Ledger suggests which member firms are worth a review, each with a short reason and no figures. Nothing is sent from the suggestion: you open Practice plans and ask yourself.
  • Unit admins see only their own units. Cohort requests reach up to ten firms at a time, with a review step first.
  • Ledger prepares your desk privately before each review. The member gets separate private questions. Each side can download its own branded brief, and neither sees the other's notes.

Recruiting desk

For firms you don't yet have.

  • Ask an advisor firm for a few figures under a recruiting NDA. The firm answers from a free account, and you get an educational estimate of its current book.
  • Accepting a reply never recomputes anything on its own. A fixed checklist and private Ledger coaching help you prepare, and the recruit gets private questions of their own.

M&A desk

Member deals and network acquisitions.

  • Pipeline turns the buy/sell view into a working list you can track. Introduce two member firms in two steps: the seller is asked first without the buyer named, and both names are shared only once both say yes.
  • Deal rooms lists member firms' own Diligence Rooms, showing only that each exists and when it was last active. No figures, notes or messages.
  • Acquisitions lets the network itself act as a buyer, for aggregators and roll-ups. Add an outside firm by CRD, or ask a member firm whether it's open to the network as a buyer. Saying no changes nothing about its membership.
  • Each step of an introduction or acquisition ask sends an email with firm and network names only, never a figure.

Succession desk

Readiness, continuity and successor introductions.

  • Cohorts come from saved report facts and owner-approved checklist summaries only, and Practice reviews can carry a succession agenda.
  • Successor introductions: an owner who opts in can be introduced to another member firm that is open to acquiring. The owner is asked first, then the other firm, and neither is named to the other until both say yes. No figures or timelines are shared.

For member firms: continuity and succession

A move card, not a banner.

  • Continuity and succession now live inside the “Close the continuity and succession gap” move card, not at the top of Moves. A five-question continuity checklist is free; succession prompts are on the Owner plan. Neither changes a figure.
  • Succession facts stay on the report's Succession readiness card, with links both ways and a count of checklist answers.

Diligence Room

One room, three types.

  • Deal Room is now called the Diligence Room. Deal diligence works as before; Recruiting diligence and Practice review are new. Old room links still work.

Your assistant can answer

Recruiting and practice requests, worked step by step.

  • A firm's own assistant can now work an incoming recruiting request or practice review step by step, the same way it works a deal request. Each request states its purpose up front.
  • A recruiting request stays covered by the recruiting NDA; a practice review relies on the member's existing network consent. The firm still approves every share and decline.
  • An assistant never accepts a suggested move or a professional's offer on the owner's behalf.

Public pages

Every desk and room type, explained.

  • The Diligence Room demo page now covers all three ways to use the room, not just buying and selling. It says who sits in each seat, which plan it needs, and that recruits and member firms join free.
  • The harness page explains the loops between firms and networks, and the agent-to-agent page covers recruiting and practice reviews, including what an assistant may never do for the owner.
  • The Network and Ledger pages now cover succession and continuity.

v1.0.34

ValuationMCP has introduced Ledger fills for every move and Ledger for NetworkNovember 11, 2026

Ledger now fills the blanks in your saved documents for every move, using the app's own calculations, and leaves each one for you to review. A saved document can turn into checklist steps, repeating tasks and next drafts, and questions for your professionals get Ledger's take. Only you can approve a fill. Network accounts can now ask Ledger about network-level data. The Value impact view, Network pages and the six lenses examples are clearer. Engine 1.8.0 unchanged.

Ledger fills and reviews

The app's figures, your decision.

  • Let Ledger fill this now works on every move, not just the household book. Each blank is filled from a calculation tied to that move or from your saved valuation, and every fill waits for your review.
  • Ledger fills your document in place and keeps your wording. He never rewrites it, and no brackets are left behind. Suggestions are listed in order of how reliable their source is.
  • Only the owner can approve a fill or enter a value, and the database enforces this. Every decision is recorded and can't be edited afterwards.
  • Pieces about client demographics and similar moves can now be labeled Policy or procedure, so saved pieces show the right type.

From document to work

A saved piece becomes the next steps.

  • Each saved piece has a From this document card that proposes checklist steps, repeating tasks and next drafts. Nothing is added until you accept it.
  • Repeating tasks are proposed only for lines that describe recurring work, not every date in the document.
  • Questions for your professionals now show Ledger's take on each question. His answers are saved, earlier ones are never overwritten, and he never states a figure.

Ledger for Network

Ask about the whole network.

  • Ledger now works in Network accounts. Ask him about network-level data, such as health, succession, moves and coverage, without opening a firm's workbench first. He answers from saved outputs only.
  • In the fictional demo network, Ledger can also send nudges and clear his chat. No emails go to the demo's fictional firms.

Tray and moves

Easier to read, fewer surprises.

  • In the tray, Value impact now shows a straight line from today's fair estimate to the value after the move, zoomed in, with clear labels. Benchmarks and How the score is set are easier to read.
  • On Move the number, the value impact chart shows every driver by default, with distinct colors and pills under the chart that turn each one on or off.
  • Ask Ledger now shows the newest exchange at the top, so you can see he is working without scrolling.
  • Setting dates and frequency and then assigning Ledger no longer clears the dates you entered. Acting on a move no longer fails quietly; if the AI service can't reply, you see an error and can try again.
  • Knowledge graph hover labels now sit at the tip of the pointer. Skill cards no longer show the ready-to-send prompt.

Networks page and Network views

Clearer charts, less scrolling.

  • On the health chart, clicking a firm opens a small card inside the chart, like the succession example, instead of a readout below it.
  • The health and succession charts now have two-row keys with small pictures of each mark, on the Networks page and in the Network app.
  • In Six lenses on every practice, picking a lens shrinks the cards into one row of buttons with the example right below, so you don't have to scroll to see it.

v1.0.33

ValuationMCP has introduced the growth loop and a deeper, more capable LedgerNovember 8, 2026

On the organic-growth move, a growth loop asks for quarterly flows first, separates net new assets from market change, and drafts plan steps you apply. A single stated growth rate is a rate, not flows, and is never treated as evidence. Ledger now explains what is behind a number, writes finished pieces you can save and export, and can be asked to do a task or work a move on a schedule, with every figure inserted by the app. Engine 1.8.0; eighteen MCP tools.

Growth loop

Flows first, never the market.

  • The growth check runs on the organic-growth move from that move's Data tab, with its status, steps and next re-check on the Plan tab too.
  • When no flows are on file, the first card says so and lists where to get them, in order: a connected portfolio accounting server, then ones you can connect in the app or through your assistant, then a flows export upload with a template.
  • A single stated growth rate is labeled a rate, not flows. It never counts as flows evidence, and with no flows a hosted run drafts nothing else.
  • Once you pull or upload quarterly flows, it shows net flows (contributions minus withdrawals), new and lost households, the organic growth rate and how much of the change in assets came from flows versus the market.
  • Market change and acquired assets are never counted as growth, and flows are never forecast. It drafts plan steps you apply; nothing is added until you click Apply.
  • It re-checks on the move's cadence and only ever proposes a fresh pull. Flow rows stay in your browser; only the totals you tick are kept.
  • The engine still uses organic growth only as a quality score in the multiplier. No figure was changed.

Growth read and send confirmation

What the flows show, before any step.

  • The growth check now leads with a read of your flows: whether growth was mostly organic, mixed or mostly market, labeled as a practice-management read, not advice.
  • It shows how much of the change in assets came from flows versus the market, how many quarters had positive flows, how flows held up in quarters when the market fell, and whether recent quarters are running ahead of or behind earlier ones.
  • When your upload has no starting assets, the organic growth rate uses the assets from your latest saved run, labeled with its date, and is shown next to the rate stated on the move. A stated rate is still never used as flows.
  • Every figure is computed by code. Ledger's short summary contains no figures, and each drafted step names the finding it acts on. Steps that only collect data you already sent are dropped, and data gaps sit in a separate list that Apply never adds to your plan.
  • The send buttons for the growth, margin and billing checks now show Sending, then Sent with the time, bring the check into view and show an inline error with Retry if a send fails.

Ledger's diagnosis, written pieces and export

The engine's figures, explained and put to work.

  • Ledger can now show figures in his answers. He names a fact and the app inserts the engine's exact value, with its label and source on hover. He still never writes or calculates a number himself: any sentence where he types a figure is removed.
  • Each answer now reads your saved run's inputs (revenue, EBITDA, AUM, clients, growth rate), ratios worked out by the app with their formulas shown (revenue and AUM per client, margin, expense ratio, revenue on AUM, uplift per score point) and where your AUM and client count sit against Form ADV peer quartiles. When no peer data exists for a driver, he says so.
  • Ask why, what is driving a number or how it compares, and Ledger answers with a What's behind this section: what the number is, what it is worth on the fair estimate, the likely causes and how to check each one.
  • Ask him to draft, write or build something, and he writes the finished piece instead of a checklist: talk tracks, fee-change memos, client letters, outreach sequences, agendas, transition plans and briefs for your CPA, attorney or insurance advisor, chosen to fit the move.
  • Client-facing pieces carry a banner, Draft for your compliance review before use, and the app never sends them. Save a piece to its move, edit it later (the previous version is kept), copy it as markdown or text, or download it as a branded PDF or Word file.
  • Pieces can include charts the app draws from your saved figures: value range, revenue to EBITDA, uplift by move and AUM against peers. Ledger can only ask for a chart by name; he never supplies its values.
  • Ledger now runs on Claude Opus 5.5 for chat, Explain and the deal room, and on Claude Sonnet 5 for the agent loops and the hosted agent. Engine 1.8.0 and the eighteen MCP tools are unchanged.

ValuationMCP has introduced the Put Ledger to work page

Agentic support for your number and every move, on one public page.

  • The new Put Ledger to work page shows how he explains a valuation, works each of the ten moves, does the legwork on request or on a schedule, writes finished pieces and turns down what he should never do.
  • Its answers and pieces are real Ledger outputs recorded on the fictional Alder Ridge demo firm, replayed without any live AI call. Every figure was inserted by the app from that firm's saved valuation.
  • Both sample pieces download as the same branded PDF and Word files the app builds. Both Put Ledger to work pop-ups now link to the page, which appears under What you can do in the Product menu.
  • From the All view, Ledger's pieces can now be saved to a move you choose. When the fact check removes every sentence of an answer, the reply now says so instead of showing nothing.

Asking Ledger to do the work

From a checklist to a finished piece, on demand or on a schedule.

  • Ask Ledger what to do first now asks him for you: the panel opens and he answers straight away, naming the move to start with, what it is worth, why it comes before the others and the first step to take this week.
  • Ledger always answers the question itself. He never replies with only a pointer or a code name, and any workflow he suggests shows as a plain-language card with a short reason.
  • Every task on a move's checklist now opens up, with an Ask Ledger to do this button. He writes the finished work as a piece you can save, says plainly what your team still has to do, and never marks the task done himself.
  • When a move is assigned to Ledger on a schedule, he checks in on his own at the move's cadence (weekly, monthly or quarterly; monthly unless you change it), taking the next open task at each check-in and leaving the result for you to apply. Pausing or revoking the assignment stops further runs.
  • The assignment choices now say who drives the work: Ledger, when you ask (he answers only when you open Ask Ledger) or Ledger on a schedule (he works the move at each check-in and leaves cards on the Plan tab).
  • Each move's panel tabs now read Plan, Data, Value impact and Ask Ledger, in that order. Value impact is the tab that was called Math.

v1.0.32

ValuationMCP has introduced deal loops over A2A, a billing loop and LedgerNovember 7, 2026

The buyer's loop proposes requests and diligence; the target's loop drafts responses from its own record. A person approves every send, share and accept. On the revenue-per-client and recurring-revenue moves, a billing loop reads the household billing you pull or upload and drafts plan steps you apply. Both sides of the Moves page are rebuilt around what to work on next: the selling board leads with your total opportunity and what is in motion, the buying board leads with the upside after close and keeps your walk-away ceiling separate from it. Engine 1.8.0 and eighteen MCP tools remain unchanged.

Deal diligence loop

Requests and responses, not open access.

  • The buyer's deal loop finds missing information and proposes data requests and diligence for the deal room. The buyer never pulls from a target's systems.
  • The target's deal loop drafts responses from its own record, showing the requested fields and sources. Every send, share and accept waits for a person.
  • Accepted figures are labeled target-provided and retain the source details the target chose to share.
  • Each message is logged once in an append-only step record. Incoming messages only become cards, never instructions or actions.
  • Both loops stop when the deal room closes or the request is revoked or expires. Nothing is drafted or shared until an NDA is in place.
  • With no NDA, the target loop waits and drafts and shares nothing. Agents never sign or waive the NDA.

Billing loop

Totals only, and nothing changes without your click.

  • On the revenue-per-client and recurring-revenue moves, you can run a billing check from the Billing panel. It reads the household billing you pull or upload.
  • The check shows the effective rate, fee mix or recurring share for that move.
  • It drafts plan steps for that move. They are added only when you click Apply.
  • Re-checks follow the move's cadence and only propose a fresh pull. The loop never changes a fee, contacts a client or computes a valuation.
  • Billing rows are never stored or logged; only the totals the owner sends are kept.

Meet Ledger

A face for the assistant inside the app.

  • ValuationMCP has introduced Ledger, the face of its in-app assistant: an owl in orange glasses who keeps an eye on your numbers.
  • Ledger appears on the header button, the assistant panel and each reply, on Ask and Explain buttons, on co-pilot assignments, and on the checks and cards it drafts for your approval.
  • A short introduction to Ledger appears the first time you open your valuation, and you can reopen it from the assistant panel.
  • Ledger's voice is short and plain. Every existing rule still applies: figures come only from the engine, and nothing changes until you approve it.

Moves board

Both sides rebuilt around what to work on next.

  • The selling board opens with your total opportunity at display size, the fair estimate it sits on top of and the date that estimate was recorded.
  • An In motion card shows how much of that total is planned or in progress, with a legend for in progress, planned, not started and verified on file, and the date of the next re-run.
  • A value bar draws each priced move at a width set by its own uplift, largest first. Choosing a segment jumps to that move.
  • A This week panel holds up to three tiles: your check-in items, the largest move in progress, a planned move waiting on its owner, or the moves still without an owner.
  • Every move is one full-width row with its figure, its facts and consequence, and one button that follows its status: Start this move, Open the plan, Log progress, Re-run to verify, or See the proof.
  • The list defaults to Largest first, with a toggle to order it for your goal and filters for All, In motion, Business, Paperwork and Data.
  • The buying board opens with the upside you could create after you close, beside a card that keeps your walk-away ceiling marked on its range and shows the upside as a stretch beyond it, never as a higher price.
  • A diligence panel names what diligence still needs from the target, one tile per item, each tagged engine default, missing or published median.
  • Where you'd create value after close gives each factor its figure, a plain diligence question, the math inline under the row and a link to why that score.
  • No figure was repriced: every amount on both pages is the same engine row it was before, and deep links, the Free one-move gate and the PDF export behave as they did.

v1.0.31

ValuationMCP has introduced governed agent loopsNovember 6, 2026

The work ValuationMCP does on your behalf now runs as governed agent loops: every step is logged once, every tool has a permission tier, and anything that changes your record waits for your click. This release also brings move progress and planning tools and a new Agent harness page. Every figure is unchanged; engine 1.8.0, still eighteen tools.

Read about the agent harness

Governed agent loops

Each loop reads, proposes and waits for you.

  • A data check runs before every valuation, in the workbench and over MCP. It finds missing, unverified, inconsistent, other-currency, unclear and default inputs and asks one plain question about each. With nothing to fix, the valuation records straight away; items that would change the result open first, with Compute anyway (provisional) and Cancel.
  • Over MCP, compute_valuation runs the same check: blocking items come back as questions instead of a valuation, unless the owner chooses to acknowledge them and save a provisional run. validate_practice runs the check on its own.
  • The operating-margin move has its own loop. On the move's Data tab, pull your P&L or upload an export to see each expense category as a share of revenue and your operating margin against your size tier's median. Plan steps are drafted for your largest categories and added only when you click Apply.
  • Margin re-checks run on the move's own cadence and only propose a fresh pull. Firms in a network get them too when the member has shared full detail.
  • The client retention, concentration and demographics moves have a book check. Pull your household book from a connected CRM or upload an export; the results are worked out in your browser and only that move's totals reach ValuationMCP, never household names. Plan steps are drafted for that move and added only when you click Apply, with re-checks on the move's cadence. Retention still changes only the quality score, and the revenue that left with departed households stays a labeled fact from your export, never a forecast.
  • The ValuationMCP Agent can now be hosted by ValuationMCP: a third choice when you add an agent co-pilot, beside your own assistant and the in-app assistant. It works the next open step on the move's cadence, or when you press Run now, and leaves every change as a card you apply or dismiss. On the close-data-gaps move it runs the data check and leaves its questions for you; on the three client book moves it uses the book check and drafts only from totals you already sent. Pause and Revoke are on the Plan tab. It never pulls or computes on its own, never sends email and never changes a figure.
  • On a buying target, missing figures are asked for through a data request under the NDA rather than pulled from the target's systems.
  • Pulls are read-only and only start when you start them.
  • Hosted runs are included on the Owner, Deal and Network plans, with a monthly cap per owner and a daily cap on Run now. When a hosted run leaves a card, the move shows a Hosted agent chip on the board and the card appears on the Plan tab and in your check-in list.
  • After you apply drafted steps, the loop shows that they are on your plan, and steps already on the checklist are not added again.
  • Over MCP, list_practices now shows when a move is hosted by ValuationMCP, with its last and next run, whether it is paused and how many cards are waiting. The server guide no longer asks assistants to run validate_practice before every compute, because compute_valuation runs the check itself.

Move progress and planning

See where each move stands and shape its checklist.

  • Add a task to any move's checklist, or ask the Assistant to suggest more. Suggestions are added only when you press Apply to plan, and tasks already on the list are not added twice.
  • A move keeps its human owner, with an optional agent co-pilot for drafts, analysis and follow-through. The person you assign can run their own co-pilot in Claude or ChatGPT; it works for them, drafts updates they post, and never posts, emails or contacts anyone.

For developers

How the loops are governed.

  • A new Agent harness page under Developers explains the checkpoints, the four permission tiers, a loop step by step and the four loops running today. Its opening illustration shows each tool call being sorted into its permission tier and written once to the run record.

Elsewhere

  • The homepage now notes that ValuationMCP comes from the team that built Truelytics.

v1.0.30

Your focus, six pages, the whole workbenchNovember 4, 2026

ValuationMCP now asks what brings you to each firm and remembers it, splits the valuation workbench into six pages, puts what matters for your focus first, and opens the whole signed-in workbench to Free accounts. Order and emphasis change; every figure is unchanged; engine 1.8.0, still eighteen tools.

What brings you here

Context only, never a number.

  • ValuationMCP asks once what brings you to each firm (grow its value, plan succession, sell, buy, finance, check a buy-sell agreement or compare) and when.
  • Your focus shows on Overview with a Change link, for selling and buying firms alike.
  • Your assistant asks the same two questions before its first valuation of a firm, and never again once a focus is on file.
  • Your focus stays private: it never reaches a network, a deal room, a data request or any counterparty.

One workbench, six pages

Same figures, easier to find.

  • ValuationMCP keeps one live workbench across the pages, so an edit on Data shows on Report and Moves before you save.
  • Old links keep working: a saved-run link opens Report, and a move link opens Moves with the assistant tray on that move.
  • The PDF export lives on Report and carries the same content as before.
  • ValuationMCP puts Save to practice and Record valuation on Data, keeps one Record valuation button beside the Report headline, and shows the unsaved-changes tag on every page.
  • Public filings on your firm now read last on Data, after Inputs and Book quality.

What comes first

Order only, never a number.

  • ValuationMCP opens Overview on a card chosen by your focus: your sale, succession readiness, what a lender reads, your buy-sell, how you compare, where the value is, or your ceiling.
  • ValuationMCP suggests the skills that fit your focus first, each with a one-line reason, and your assistant suggests the same ones.
  • ValuationMCP orders your moves by your focus and timeline, with a reason on each and a note on how the order is chosen. What each move is worth does not change.
  • ValuationMCP names the one data gap that matters most for your focus, and you can now add an optional second goal.
  • ValuationMCP shows your total opportunity from Moves on your Overview card, with a See all moves button and your priority move card. For succession it adds up only the moves for your succession. Moves can overlap, so the total is not a combined valuation.
  • ValuationMCP now suggests the succession and buy-sell reads for a firm you are buying when your focus there is succession or a buy-sell check.
  • ValuationMCP's What-next pills now offer Run Skill, or Run Artifact for the skills that hand back a living page, and open the prompt in Claude or ChatGPT instead of only copying it.

Free plan: see the whole workbench

Show the section, gate the action.

  • ValuationMCP opens the whole signed-in workbench to Free accounts, exactly as an Owner sees it. Actions that belong to a paid plan show an upgrade card in place instead of hiding the section: sell-side actions point to Owner, buy-side actions to Deal.
  • ValuationMCP records a Free account's first valuation from the workbench or an assistant and keeps it on file with its full report.
  • The blend detail, Move the number board, move cards and the Math tab stay open to read on Free.
  • Public filings on your firm (Form ADV, Form 5500, Form 13F and stated fees) stay on Owner, Deal and Network.
  • Owner adds recording more valuations, the AI Assistant and Explain this valuation, method weighting, blend adjustments and planning, assigning and tracking moves.
  • In Claude and ChatGPT, ValuationMCP gives Free accounts one run of each workflow skill; a repeat load within a day counts as the same run.
  • Firms invited by a Deal buyer can re-save their own valuation while the request is open.

v1.0.29

ValuationMCP has introduced Network InsightsNovember 2, 2026

Network Insights rolls members' saved valuations into six lenses in the Network console (Health and wellness, Optimization, Succession, Buy/sell, Engagement and Data supply) each its own read-only dashboard over saved runs, plus an Overview of headline tiles and five action queues. Engine 1.8.0 unchanged; still eighteen tools.

Six lenses on the same firms

One question per dashboard.

  • Health and wellness plots each firm's saved book quality against its saved data confidence, with unit rollups and the weakest factor.
  • Optimization sums the uplift saved on each move by its firm's latest valuation (straight sums, never recomputed) and shows where moves stall.
  • Succession lines every sell-lane owner up by years to retirement and readiness, and marks the danger zone: five years or fewer with no successor.
  • Buy/sell lists likely sellers and in-network buyers, the saved AUM in play, and the furthest live conversation each seller has.
  • Engagement bands members by last successful activity and counts the channels firms use.
  • Data supply shows where every value behind a valuation came from: system-pulled, owner-stated or benchmark-derived.

Overview starts the week

Tiles and queues.

  • Headline tiles give each lens one number, read from the same saved-only rollups the dashboards use.
  • Five action queues (due for re-value, low data confidence, unacknowledged data gaps, never valued, valued under an outdated policy) each link to the page that fixes it, and anyone who manages members can nudge a practice from the list.
  • The fair-value distribution across the network lives on Overview, suppressed under five practices.

Where it lives

The Network console.

  • Insights sits in the Network console beside Overview, Members, Reports, Governance and Settings.
  • Every figure is copied from a saved valuation, its input snapshot or the results-only rollup; groups under five firms show no counts or dollar rollups.

v1.0.28

Every data request now opens a shared deal roomNovember 1, 2026

Sending a target a data request opens a shared deal room between the two firms, and the whole workflow (NDA, figures and questions) now lives there for both sides. The system-generated NDA became a real agreement both parties opt into, and a target with a recorded valuation shares straight from it. Engine 1.8.0 unchanged; server 1.36.0; still eighteen tools.

The request opens the room

One workflow, one surface, both firms.

  • Sending a data request now opens a shared deal room between the two firms automatically. The request, the NDA and the figures all live in the room from the first step, on the web or through an agent.
  • A target without an account is asked to open a free one first, then joins the buyer's room from the email link, both firms need an account.
  • In-app delivery is now the default whenever the target has verified its firm and opted in, so the request lands in the Requests inbox. The notice email that goes with it names no figures.
  • Every step alerts both parties: the Requests page lists what's waiting on you, the header badge counts it, and an email goes out at each step. Step emails carry names and dates, never figures.
  • The workbench request card and the Requests inbox both offer an Open deal room shortcut to the room.

The standard NDA is a real agreement

Opt in or bring your own.

  • The old “pending legal review” draft is replaced by a system-generated NDA (template 1.0). Both parties read it, accept it in their own seat, and the acceptance and signed PDF are stored with the room like an uploaded agreement.
  • Either side can instead upload its own NDA for the other to sign; the two paths sit as Option 1 and Option 2 in the room's NDA step.
  • After you accept, the room tells you exactly what happens next, and the waive option moved into its own clearly labeled box with separate consent text.

Sharing data, the easy way

Start from what's recorded.

  • If the target has a recorded valuation, the request offers Share from your recorded valuation first: a checklist of figures, each with its source line, sent only after review.
  • With no recorded valuation, the room prompts Run your valuation first, and the survey stays as a collapsed last resort.
  • Figures shared through the request now show as their own darker block for both sides, separate from the firm's filed Form ADV facts.
  • The buyer-hosted room reads cleaner: filed facts four across and the What diligence replaces list on its own row beneath the shared figures. The optional share-more block stays switched off while it's under review.

Agents get the same room

A2A parity with the app.

  • A request created by an agent opens the same shared room and returns its link. list_practices now shows each target's NDA status and deal room link, and agents default to in-app delivery when the target is reachable.
  • Public A2A pages and the Agent Card now describe the shared room, the standard NDA and the both-firms-need-an-account rule.

Smaller things owners asked for

Shipped alongside the room work.

  • A Form 5500 search that finds no filings records “No 5500 under this CRD” on the workbench, and the pill survives a reload.
  • The “Let buyers send information requests” toggle now sits on each selling firm's card next to its verification status.
  • Skills & artifacts prompts are personalized to your firm instead of a generic example.
  • The welcome popup leads with firm setup, then connect, and the Connect page no longer scrolls sideways on step 3.

v1.0.27

Self-building artifacts, a real agent assignment and a rebuilt Connect pageOctober 30, 2026

A skill that produces an artifact builds one the first time it runs and carries a one-click refresh. Naming the ValuationMCP Agent as a move's owner creates a real assignment, with a playbook for that category and a schedule you choose. Settings → Connect is rebuilt around the MCP connection, Claude and ChatGPT deep-link with their brand marks. Engine 1.8.0 unchanged; server 1.35.0; still eighteen tools.

Artifacts build themselves, and refresh

You no longer have to ask for the document, or re-prompt to update it.

  • The first time a skill that produces an artifact runs in a chat, the artifact is built without being asked, then you're asked whether to keep it. Saying no leaves it as a preview; a later run updates the same artifact rather than prompting again.
  • Every artifact carries a Refresh with latest data button in its header. It re-runs the skill's tools for the same subject and replaces every figure with what the tools return now, updating the generated date. It never edits a figure inside the document.
  • Artifact accent text now uses the same orange as the headline figure (the logo, the firm name, links and values) so the document reads in one voice. Green stays for chart bars and lines.

The agent that works a move

Naming the ValuationMCP Agent now actually assigns the work.

  • Naming the ValuationMCP Agent as a move's owner (in Claude, ChatGPT or the app) creates a real assignment, and the Plan tab shows the agent's identity (assistant · cadence · last report) instead of You.
  • Each of the ten move categories has a playbook: the tasks to establish, the data to analyze, the draft documents to generate, and the review boundaries the agent won't cross. A model can never state or alter a figure.
  • The agent asks how often to run the work (daily, weekly, monthly or quarterly) and needs an explicit time, time zone and your approval before anything is scheduled outside ValuationMCP.
  • Agent assignments don't expire on a person's clock: they run until you confirm done or revoke, with a year as a backstop and no expiry date on the Plan tab. A report arriving for a move with no active agent assignment is ignored rather than refused; revoked assignments are still refused.
  • The tray's owner field no longer cuts off a name.

Settings → Connect, rebuilt

One path first, the rest optional.

  • Step one is the MCP connection: the address, a Copy button and setup instructions, open by default. Step two, Connect your data, is now marked Optional.
  • The connect list shows only the platforms the app can sign you into directly, with a note that the set is limited but growing. Your own filing servers (ADVmcp, 5500mcp and 13Fmcp) read as connected with no login, because they need none.
  • Claude and ChatGPT are whole-card links carrying their brand marks, which flip between dark and light mode. Claude opens its connector form with the address filled in; ChatGPT opens its plugins page. Cursor and Codex were removed, their setup isn't friendly enough to sit on this surface.
  • Step tabs and the action bar stay pinned so Next never needs a scroll, and Continue now reads Next.
  • Archived firms moved from Connect to the Firms page; Network operators find Admin under Settings and no longer see their own firm listed as a member; the footer's Product list gained a Demo link.

Smaller things owners asked for

Asked for in the last two weeks, shipped in this one.

  • The coverage percentage expands to show every field's value or presence, its weight, the source and date behind it, and the recorded receipts.
  • The Workbench offers Run Valuation for a firm with no saved run and Rerun Valuation once one exists, opening your preferred assistant with a firm-specific prompt you can review and copy. Explain only appears when there's a valuation to explain.
  • Google sign-in always shows the account chooser, so a second account is one click away.
  • Pages load faster: skill icons are small local images instead of large remote ones, and repeated homepage and skills reads reuse recent data.
  • The Workbench move assistant's replies stopped failing; the answer format it asked the AI service for used a constraint the service rejects.

v1.0.26

Assign a move: to a person or the ValuationMCP AgentOctober 29, 2026

Owners can hand one move to a partner, employee or outside professional, or to the ValuationMCP Agent: your own Claude, ChatGPT or other assistant that works the move on a schedule and reports back. The tray's Plan tab and the assistant share one review form and one send; a person gets a private link and reports progress without an account, the agent reports through your MCP connection, and the owner confirms when it's done. Engine 1.8.0 unchanged; server 1.31.0; still eighteen tools.

Assign this move

Action cards in the tray, starting with one.

  • The Plan tab's Owner field reads You or Assign someone…: name, email, role, an optional note, whether to include this move's figures (off by default) and whether to send reminders. Review shows exactly what goes out before Send.
  • The assistant can suggest an Assign this move card under a reply when you name someone or the work belongs to a partner, employee or professional. It suggests only the move and a reason in words (never a person, email, date or figure) and Review opens the same form.
  • The assignee's email links to a private page for that one move: the brief, the checklist, status, notes, I'm blocked and Request a new date, a .md download and a copy for their own assistant. With figures off, the brief and emails carry no valuation figure and don't mention a valuation or a sale.
  • Ticks and In progress update your plan; Done is reported by the assignee and becomes done only when you confirm it. Date requests change nothing until you accept. Every update appears in the tray chat and the Plan tab, and check-ins flag reported work, blockers, date requests and unopened assignments.
  • Reminders follow the move's cadence on the existing daily check-in, never within a week of sending, and stop when the work is reported done, revoked, expired or the assignee stops them in one click. Revoke and Resend take effect at once.

ValuationMCP Agent

Your own assistant, on a schedule.

  • The assign form offers A person or ValuationMCP Agent on every move. For the agent you pick the assistant, add instructions, and choose whether to include figures (on by default; it is your own assistant, signed in as you). No email is sent.
  • You get a schedule kit to copy, or open prefilled in Claude or ChatGPT. Your assistant turns it into its own scheduled task at the move's cadence, and between runs answers as your practice-management, M&A and succession consultant for that move.
  • Each run reports through ValuationMCP: ticks and notes apply, blockers and date requests wait for you, and done is reported. You confirm it. The agent never changes a figure or a metric in a scheduled run and never contacts a client.
  • The Plan tab shows ValuationMCP Agent (assistant) · cadence · last report, with Show the kit, Revoke and Confirm done. Revoking stops the next report. Check-ins flag an agent that has not reported for two periods.

v1.0.25

Plan my moves, connected data and valuation guidanceOctober 28, 2026

The workbench's "How to move the number" becomes Your plan: the same priced moves, now in three zones (Start here, Everything holding the number back, Already working for you) each with its kind, the fixed worth sentence, arithmetic from your inputs, a first step, how you'll know and the benchmark behind it. Owners can plan a move, tick its checklist and set an owner, due date and cadence; the plan lives on the practice and every status change is recorded. Engine 1.8.0 unchanged. Every figure is identical; server 1.30.0; still eighteen tools.

Your plan

The driver block, organized as moves an owner can plan, track and prove.

  • The tray assistant now has focused expertise for every move category. Ask about this starts an answer using the selected billing or client-book section and its exact analysis totals, including synthetic connected sources; follow-ups retain that context only for the same firm and move. Household rows stay private and valuation math is unchanged.
  • Your firms (picking, adding, renaming and archiving practices, and the selected firm's data and data requests) now lives on the Your firm(s) page under Settings. The Connect page keeps the agent connection steps, coverage and the archived list.
  • An isolated integration test now runs the full demo reset, checks both generated NDA PDFs and their recorded document hashes, verifies all seven Network sell firms have connected analyses, and ensures a failed document upload cannot report success. Test runs do not change live demo accounts.
  • Demo reset sets up the synthetic CRM and billing connection and both analysis reports for all seven sell-side Network firms, as well as the standalone sell demos. Reset reports success only after every sell demo is set up; valuations and buy targets are unchanged.
  • Each sell demo gets Cottonwood's one-click synthetic billing and client-book pulls as soon as its firm is created, even if a later reset step fails. Source-loading errors are shown rather than mistaken for no connections.
  • Demo reset clears abandoned impersonation records only when their sign-in session no longer exists, while still blocking resets during a live demo session.
  • Demo reset uses a self-contained document generator for fictional NDAs and checks it before changing demo data, avoiding a PDF helper import failure partway through reset.
  • The side tray can expand to the left for wider client book and fee analysis tables. The expand icon beside Close restores the previous width on a second click; switching widths keeps the current results and selections intact. On phones the tray already fills the screen.
  • Three zones in a fixed order, in the server's order inside each and never re-sorted. The top Start here card is still the report's first holding bullet.
  • Each card reads: kind (Paperwork, Data or Business), effort and horizon; the fact and its consequence; "About … on the fair estimate at a score of 85, every other factor held still"; a second line of arithmetic from your inputs, shown only when the raw metric is on file; Start here; How we'll know; and the benchmark, with a currency note when your figures aren't in USD.
  • Plan this move, Show the math and Ask an agent sit on every card. Planning opens a checklist with owner, due date, cadence and notes; the first time a move is planned, the fair estimate it started from is captured once.
  • Ask an agent now answers on the card for Owner plans and above: the move assistant reads your last saved valuation and the live Plan my moves skill, writes words only, any line stating a number is removed, and cites figures by the engine's own displays. Apply to plan adds its proposed steps after yours and sets only the owner, due date or cadence it proposed. Use your own assistant keeps the copyable prompt.
  • A progress strip shows the baseline, how many moves are planned, in progress and done, and the next re-run 90 days after the last recorded run; a standing list says what won't move the number.
  • Free accounts can plan one move; Owner and above plan every move.

Plan my moves: for assistants

Additive only: no new tool, no changed input.

  • compute_valuation rows gain kind, proof, benchmark, businessEffect and firstStep, plus a wontMove list. list_practices(practice_id) returns movePlan; upsert_practice accepts an optional moves[] written only after the owner confirms.
  • value-driver-playbook 3.0.0 is now titled Plan my moves (same slug); presentation-contract 2.6.0 adds the move vocabulary; four skills read the plan before suggesting work on a driver.

Prove a move

A move marked done can now be proved: when a saved valuation shows its measure moved on file, the server marks it verified with the before and after reading and where the figure came from. A check-in lists what's overdue, due soon, stalled or waiting for proof, and Owner plans can choose to get it by email. Two new assistant workflows prove moves and run check-ins, and the succession workflow now runs a short interview. Engine 1.8.0 unchanged. Every figure is identical; server 1.29.0; still eighteen tools.

    Proof and check-ins

    Verified only by the server, only from figures on file.

    • On a saved valuation, each move you marked done is compared with the run it started from. If its measure rose (and, for book-quality moves, the figure is on file rather than a placeholder) the move becomes Verified, with the reading before and after and its source. If not, it stays Done and says so: confirm the new figure and re-run.
    • What changed lists the moves verified on that run, and never credits them with the change in value.
    • A Check-in card at the top of Your plan, and on your saved valuations, lists overdue and due-soon moves, moves with no step ticked in thirty days, moves waiting for proof and whether a re-run is due.
    • Owner plans and above can turn on check-in emails per practice, weekly, monthly or quarterly; off until you choose. Figures in the email are copied from your last saved valuation.

    Prove a move: for assistants

    Additive only: no new tool, no changed input.

    • list_practices gains a check-in note per practice; compute_valuation comparisons gain the moves verified on that run.
    • New skills prove-a-move and move-check-in; succession-readiness-gap 2.0.0 adds a structured interview; valuation-delta 1.3.0 adds the moves line.

    Pricing and fees, MCP-first

    The revenue-per-household move gets a fee analysis: realized yield, households below a fee floor, discounts against your stated schedule and revenue-only what-ifs. It starts where your billing data already lives: a connected billing or portfolio server, then lists the servers you could connect, and only then offers an export upload. Household rows are never stored. Engine 1.8.0 unchanged. Every figure is identical; server 1.30.0; still eighteen tools.

      Where the data lives

      Connected servers first, connectable servers next, an export last.

      • Connected servers that can supply household billing come first: pull from one, map its report once, and next quarter's pull is one click. Rows pass through ValuationMCP's server to reach your browser and are not stored; only your mapping is saved.
      • Servers you could connect are listed with their kind, hosting and sign-in. Remote servers with a published endpoint connect in the app; the others are set up in your own assistant, with the steps and a copyable prompt.
      • Saved figures record the server they came from, or Billing export (CSV).

      Pricing and fees, MCP-first, for assistants

      Additive only: no new tool, no changed input.

      • list_data_requirements gains a datasets block: household billing's columns and the catalog servers that can supply it, with their connect steps.
      • pricing-and-fees 1.1.0 pulls from a connected server first, then helps the owner connect one, and only then asks for an export. Rows stay in the client; aggregates only.

      Client book

      The retention, concentration and demographics moves share one client book analysis: household retention over the last year, an at-risk review list, the top ten's dependence on single advisors, the age and life-stage mix and next-generation coverage. It starts with a connected CRM, then lists the CRMs you could connect, and only then offers an export upload. Revenue that left with departed households is a fact about last year, never a forecast. Engine 1.8.0 unchanged. Every figure is identical; server 1.30.0; still eighteen tools.

        One analysis, three moves

        The card you open from comes first.

        • Run the client book analysis from the retention, concentration or demographics card. Household names stay in your browser; tables show up to two hundred rows and download in full.
        • Retention needs your departed households. Without them the panel says so and shows no rate.
        • Save the aggregates you tick: the source is recorded first. AUM saves only when you confirm your CRM's AUM is complete.
        • On the retention card, try last year's departed revenue as an unsaved what-if in the workbench.
        • Current coverage now sits above both connect paths, since your agent and the app both raise it.

        Client book: for assistants

        Additive only: no new tool, no changed input.

        • list_data_requirements' datasets block gains the client book: its columns and the CRM servers that can supply it.
        • client-book 1.0.0 serves the three moves MCP-first and drafts outreach with placeholders; value-driver-playbook 3.3.0 hands off to it.

        Billing panel

        Analysis buttons now appear only on moves that are holding the number back. Strengths stay one line, each with a chip naming where its figure came from. The fee analysis became the Billing panel: fees, recurring mix and concentration. Engine 1.8.0 unchanged; server 1.30.0; still eighteen tools.

          Where the buttons go

          One rule for every card.

          • “Analyze your billing” on the revenue per household, recurring revenue and concentration moves; concentration also keeps the client book button beside it.
          • Every line shows owner-stated, benchmark-derived or system-pulled. The first two link to “verify”, which opens the pull that can replace them.
          • The data coverage card offers “Pull from your billing system” and “Pull from your CRM”, each listing the figures it can replace.
          • The recurring mix lists commission and one-off households ranked by fees: your re-papering list.

          Agent shelf

          The workbench has one work surface: a panel on the right with Plan, Math, Data and Assistant. Move cards are compact and open it. One conversation per firm keeps the move you asked about beside each message. Engine 1.8.0 unchanged; server 1.30.0; still eighteen tools.

            One place to work

            The board is the map; the panel is the work.

            • “Open” on a card shows its plan, or the math for a strength; dataset buttons open the analysis.
            • The Assistant button asks about the whole board; replies may propose changes to several moves, each applied on its own.
            • The panel remembers its width on this device and turns fully opaque when you ask for less transparency.

            Explain this valuation

            A button under the suggested asking price opens the panel with your valuation in focus. A valuation consultant explains what the asking price, range, coverage and methods mean in plain words, citing the engine's figures verbatim, and suggests up to three next steps that open existing pages and workflows. New guided workflow: Explain my valuation 1.1.0. Engine 1.8.0 unchanged; server 1.30.0; still eighteen tools.

              Your valuation, explained

              Plain words beside the engine's own figures.

              • The consultant explains why each figure was calculated the way it was, from your own facts and the published rules behind the engine, not only what the figure is.
              • Ask what the number means, why there is a range, which methods drive it and how confident it is.
              • Next best actions link to a move on your board, the Deal room, your saved valuations, or a workflow to run in your own assistant. Nothing is changed without you.
              • These chats join the firm's conversation, tagged Valuation; choose All to see everything in time order.

              A tidier signed-in header

              The signed-in header shows the logo at the same size as the home page, with every account control on one even line.

              • Requests moves out of the menu strip into a button beside Assistant, with a small count of unread requests when there are any.
              • Assistant, Requests, the light/dark switch, your email (which opens your settings) and Sign out sit in one fixed order at one height.
              • Network accounts find Settings, Billing and Network under the same Settings menu firms use.
              • The menu strip reads Valuation, Deal room, Skills & artifacts, Settings. The Valuation menu opens Workbench and Recorded; the Settings menu opens Connect (the firm data hub, formerly Workspace), Your firm(s), Billing and Network. The signed-in button on the public pages reads Connect too.
              • Signing in now lands on the Workbench; links that name a destination (consent and request flows) still go where they point.

              Billing and Client book, easier to follow

              Each panel now says what it answers and what to do next.

              • Each panel opens with one sentence on the question it answers and one line on the next step.
              • Once a report is pulled or uploaded, the list of servers folds into one line with Change source, so the results sit directly under the heading instead of below every server you could connect.
              • Servers you could connect now sit behind a single click; only your connected servers show until you open the list.
              • The one confirmation each analysis needs comes first: the full-trailing-year check in Billing, and how your CRM's words are read in Client book, each with a short reason.
              • A server call that returns a refusal is now recorded as an error in the activity log, not as ok.

              v1.0.24

              Ask the target, then read outside inOctober 27, 2026

              Two steps where there was one: the buy lane now asks the target first and reads the public filings second, under the name Outside in, placed by plan, and hidden until the firm has a CRD. Form ADV's estimate can be recorded as a dated, append-only run, and Form 13F listed holdings arrive from 13fmcp.ai as the fourth public source, which Owner accounts can now read on their own firm. Every saved figure in the workspace reads beside its own source and date, and the workbench says what to do next and why it declined to price a firm. Three demos can be handed over with a password and an expiry, paid upgrades can be credited back to the ad click that earned them, and the weekly WealthMCP watch is narrowed to what prices a valuation. Engine 1.8.0 unchanged; server 1.27.0; still eighteen tools, so nothing new to load on reconnect.

              The buy-lane workbench, in two steps

              Two step cards, ordered by lead time, placed by plan and named for what they do.

              • The buy-lane workbench is now two cards: "Ask the target" first (the private request under NDA and the room link), then "Outside in" with the four public sources. The separate "See what a buyer sees" preview is gone; the four-source step is that read now.
              • Placement follows the plan. On Deal the two cards sit together, so the public reading is one scroll from the request; on Owner, where there is no target to ask, the same four sources appear below "What changed?". Both are hidden until the firm has a CRD, and neither title says "Step:" any more.
              • A buyer-hosted room still shares the target's filed facts and your diligence needs, never the walk-away ceiling.

              What the filings give you, and what they never will

              Two of the four sources gained real substance this release: one you can record, one you can only read.

              • Form ADV Part 1A is the first public source: filed regulatory AUM and clients, the filing date, the target's CRD and the advmcp.ai provenance line. A field the filing doesn't hold reads "not in the filing data", never zero.
              • "Record outside-in valuation" saves that estimate as a dated, append-only record, range, confidence, engine version, the filing date it came from, and the inputs used. It lives in its own table, so it never consumes the free full run, never becomes a target's latest valuation, and never stores the walk-away ceiling. It needs a CRD, is rate-limited, and re-reads the filing at save time rather than trusting the screen.
              • Form 13F via 13fmcp.ai is the fourth source. Only a filer whose adviser link is exactly the stored CRD is offered: a CRD on the cover page or an SEC file-number match attaches with one click, a name-and-ZIP match asks you to confirm. Several filers on one CRD stay separate, each with its own CIK and figures, values are never added across filers.
              • Filed figures are labeled filed; anything 13fmcp computes (the top-ten weight, the concentration measure, filed value as a share of ADV regulatory AUM) is labeled as theirs and passed through verbatim. "No 13F under this CRD" is common and never means the firm holds no equities. None of it is an engine input, coverage or confidence, and the block says so.

              Owner accounts get the outside-in read on their own firm

              The four public sources a buyer uses on a target now open on an owner's own saved firm.

              • Any paid plan that carries public evidence (Owner, Deal or Network) sees the four sources on a saved firm, where the card reads "Public filings on your firm". Private data requests, shared rooms and the walk-away ceiling stay Deal-only.
              • One rule now decides who sees filing evidence, and it was pushed down into the database guards: the plan book, the 13F filer links and the outside-in record all accept a sell-lane firm as well as a buy-lane one.
              • An Owner account with a CRD also gets its own stated Part 2A fee schedule and IAPD brochure link, without a buy-lane run first. Coverage, confidence and valuation figures are untouched either way.

              Every figure says where it came from

              The firm's data panel gained a second column: the source of each value and the date it was recorded.

              • In "View this firm's data", each value reads beside the source that supplied it and when it was recorded, so a figure can be challenged at the moment it is seen. A value you typed yourself reads as entered by hand.
              • The sample firm carries its own source records (accounting, custodian, CRM and a succession questionnaire) so the column reads the same way on a demo as on a live firm.

              The workbench says what to do next, and why it declined

              Guidance, an upgrade path, and a withheld number that explains itself.

              • "How to move the number" is a block of its own now, no longer nested inside Book quality. Its cards rank by where this firm is weakest rather than by the size of the factor (priced economics lead, the widest quality gaps follow, unpriced items last) each naming the firm's figure, the consequence and an "N pts below target" badge. The same block is captured in the workbench PDF export.
              • An Owner account that picks the Buying lane is shown "Upgrade to Deal" beside what Deal adds and the price; an account already billed here is pointed at Billing for the prorated change. The refusal holds at the server too, so no path quietly creates a buy-side target without Deal.
              • A firm the engine declines to price says so plainly: "No valuation for this firm. It isn't a retail practice", followed by the filing figure that caused it, in the workbench, the targets list and the saved-runs list.
              • A target invited to share data is pointed first at its own connected data: the free plan, then data connections, then its own assistant. The self-reported survey sits behind a closed "Alternatively, complete the survey" disclosure, and an invitation still grants nothing paid.

              Behind the scenes

              Three things that run below the product: hand-over demos, ad crediting, and a narrowed directory watch.

              • Owner, Deal and Enterprise demos can be handed over with a password, an expiry and instant revocation, signing in at the demo sign-in page rather than the ordinary one. A shared demo session never reaches platform administration, a reset is refused while one is open, and the grants and guards arrived in migration 0048.
              • A subscription that started from an ad can be credited back to the click that earned it: click identifiers held 90 days in first-party storage and attached at signup, the payment provider's own invoice amount and date as the conversion, and a credentials-guarded CSV feed at /api/public/ads/offline-conversions.csv for Google Ads to read (migration 0046). The /value landing offers both routes: the free valuation or straight to Owner, under the same disclosure and off switch.
              • A weekly check now watches the ten WealthMCP directory categories that can supply a valuation input, and reports what changed without publishing anything: the baseline was rebuilt under those categories (109 listings, 42 already in our library, none waiting), and the supported list stays the 52 chosen by hand (migration 0047).
              • Two crowded surfaces were rebuilt around one control: every Network filter now uses a shared label-over-control layout so a label can't sit on the dropdown it describes, and adding a firm is a single field, type a name or CRD and create it straight away, or search SEC filings and pick a match.

              Smaller changes

              • Each public source carries the official site icon of the service behind it (advmcp.ai, 5500mcp.ai, 13fmcp.ai) as a small link that opens in a new tab, and sources are tagged by what they do: "Feeds the valuation", "Evidence only", "Private, under NDA".
              • Both step cards ask for a CRD at the top when a firm lacks one, and the Workspace firm form now asks for it on the selling lane too; picking a filing match fills the Form ADV figures, and any figure you entered yourself still wins.
              • NDAs between firms moved from Workspace to the deal room, where the agreements are actually used.
              • When a brochure carries no applicable fee schedule, the stated-fees block shows only the link to the brochure on IAPD.
              • The footer's network line lists 13fmcp alongside advmcp and 5500mcp.

              v1.0.23

              A Network you can supervise, evidence you can collect, and email that actually arrivesOctober 26, 2026

              The Network plan gains its supervision layer: administrators review a member's records in a view-only delegated session that always states a reason and is recorded, a unit hierarchy with scoped roles, action queues and nudges, distributions that refuse to name a practice, and a versioned Network valuation policy. Four synthetic demo accounts plus ten demo members let an admin walk the product end to end without touching a real record. App email now actually goes out (receipts, deal-room invites, contact replies and Network nudges) from a branded template set on our own sending domain. The buy-lane workbench reads as one Collect data step, its plan book now finds the adviser by name instead of asking for an EIN, and a data request can be sent from your own mail client or through ValuationMCP. Advertising measurement arrived with a disclosure and an off switch. Engine 1.8.0 unchanged; server 1.26.0; still eighteen tools, so nothing new to load on reconnect.

              Delegated Access: a reason, a record, and no way to edit

              A Network administrator can look at a member's practice, never change it, and the member can see exactly who looked and why.

              • An administrator opens a view-only session on one member's records after typing a written reason (1–1,000 characters). The session is one at a time per administrator, is attributed to them, and can be ended from either side.
              • The database itself refuses writes while a delegated session is active (practices, valuations, data sources, data requests, deal rooms, agreements and firm claims are all guarded) so view-only is enforced below the app, not by a button being hidden.
              • A member's own Network access panel lists every delegated session over their records: who, when, whether it is still active, the reason, and which records were opened. They can download their own copy first, upgrade consent from saved-outputs-only to full detail, transfer existing practices deliberately (All, selected, or None), and revoke access.
              • Administrator roles are named and scoped: owner, practice manager, compliance, analyst, plus unit administrator and viewer bound to a unit subtree. Analysts receive aggregate reports only: no member details and no delegated access.
              • Every action across web, MCP and A2A lands in one append-only activity log whose rows are hash-chained to the previous entry, so an edited or deleted row is detectable. Saved valuations additionally carry a SHA-256 fingerprint of the saved output alongside the engine, index and market-data versions.
              • The Run audit section shows that fingerprint per saved valuation, the saved input snapshot behind it, whether data gaps were acknowledged at recording, and which administrator reviews touched that run.

              The Network console: hierarchy, queues, policies

              Reporting and administration split apart, and the console can now run a network rather than just describe one.

              • The Network page is organized as Overview, Members, Activity, Data access, Adoption, Run audit and Admin; legacy anchors redirect rather than dead-end. Reporting stays in Overview, policy and membership administration moved to Admin.
              • Units form a hierarchy of region, office/OSJ and branch. Practices and administrators can be bound to a unit, and a unit administrator or viewer sees only that subtree, enforced by row-level security, with the on-screen filter applied on top.
              • Practices can be assigned to units one at a time or in bulk from pasted CSV (member email or practice name, unit path). Bulk assignment previews every change before it is applied.
              • Overview carries five action queues built from each member's latest saved run: due for re-value (over 90 days), low data confidence, unacknowledged data gaps, never valued, and valued under an outdated policy.
              • A queue item can be nudged, in-app plus an email to the member, with a hard limit of one nudge per practice per seven days, recorded to the activity log. Demo members are never emailed.
              • The confidence threshold that drives the low-confidence queue is configurable per Network (default 70%).
              • Distributions over saved outputs show median, P25, P75, min, max and average within AUM bands (< $100M, $100–250M, $250–500M, $500M–1B, > $1B), and are suppressed entirely below five practices so an individual member cannot be backed out of a group.
              • A Network valuation policy is a set of method weights that must total 100. It is drafted, published, and immutable once published; members may still use their own weights, and every saved run records the policy version it used or reads as member-defined.
              • The Adoption tab reports events, saved runs, skills invoked, distinct AI clients and active members for a period (this month, last 30 or 90 days, this quarter, or a custom range) each against the equally long window before it, with a weekly trend.
              • A period can be issued as a report: the figures, the period, the engine version and who issued it are stored, so a number handed to a board is traceable to the run that produced it.
              • The activity view shows times in the viewer's own time zone, filters by member, channel, outcome and free text, and exports the filtered rows as CSV.
              • Overview analytics were rebuilt around quartile plots, method-comparison bars, range comparisons and a confidence gauge; the Members tab keeps account details and read-only review apart from those analytics.

              Demo accounts an admin can walk through

              Four plan demos and ten synthetic members, produced by the real engine, fenced off from everything real.

              • Free, Owner, Deal and Network demo accounts plus a unit-administrator demo and ten synthetic Network members, all fictional names and non-deliverable emails with no real CRDs, seeded idempotently by one admin action and resettable to a clean state.
              • Every demo valuation comes from the unchanged production engine, so what a prospect sees is what the product actually produces. Demo entitlements are recorded as their own source with no payment-provider objects behind them.
              • An administrator can fully impersonate a demo account, not a real one, with a persistent banner naming the account being viewed and a one-click end that restores administration.
              • Real billing, firm-claim confirmation and outbound contact messages are disabled inside a demo, and demo sessions exist on the web only: the same account through MCP or A2A is refused.
              • A protected flag on the account is the single source of truth for what is a demo, so demo practices, valuations and members stay out of real revenue, analytics, founding-member, trial and cohort roll-ups and out of cross-user discovery.

              Email that arrives, on our own domain

              One branded template set, sent through the platform's managed email service from notify.valuationmcp.ai.

              • Six templates share one ValuationMCP shell, logo, accent bar, the legal and HnyComb network footer: the payment receipt a buyer gets when an invoice is paid, the contact-form confirmation and the internal alert it raises, the deal-room invitation, the Network nudge, and the data request a target receives.
              • The receipt carries the amount actually paid, the line description, the invoice number, the paid date and the hosted invoice link, and is sent from the payment provider's signed notification, so it follows money, not a button press.
              • A deal-room invitation goes to the invited address only when the room is still open and its host asks for it; an opted-out address is reported rather than silently skipped, and the join link stays copyable.
              • A Network nudge lands in the member's inbox with the reason in plain language: "the latest saved valuation is more than 90 days old", and a link straight to the practice.
              • Each send carries an idempotency key, so a retried event cannot produce a second copy of the same email, and suppression and rate limits are enforced by the sending service.

              Buy lane: one Collect data step, and a plan book that finds the adviser

              The card that used to lead with a request button now reads as a three-source evidence step.

              • The buy-lane card is now headed Step: Collect data: Gather evidence on the target, with three numbered sources side by side: request data from the target (Private), the retirement plan book from Form 5500 (Public), and stated fees from Form ADV Part 2A (Public). Each collapses independently and the layout is remembered per person per target; a divider then offers a shared room.
              • The plan book no longer asks you to type an EIN. You search the adviser by name, prefilled with the target's name, and 5500mcp returns provider identities grouped by EIN, one card each, with the plans, participants and sample sponsors it appears on.
              • A card whose EIN matches the target's Form ADV filing is badged "Matches Form ADV EIN", but nothing is ever auto-selected: the book attaches only when you click "This is the adviser" on the card you confirm.
              • Confirming attaches every plan filed for that EIN, fetched page by page, and any plan can be removed individually afterwards. Where the filings hold no value, the book reads "not filed", never a zero.
              • The old EIN entry stays as a fallback behind "I have the EIN", and the whole flow switches itself off politely if 5500mcp's search capability is not reported: it says provider search is unavailable and asks for the EIN instead.
              • Fees per participant and fees as a percentage of assets are labeled derived, and the percentage was corrected: 5500mcp already returns it as a percentage, so the extra scaling that showed 111% for a 1.11% schedule is gone.
              • The five coverage caveats are unchanged, collapsed under "Coverage limits", and the book still says plainly that it does not change the valuation. It remains filed evidence: never an engine input, never coverage, never confidence.
              • Stated fees now link to the firm's Part 2A brochures on the SEC's IAPD site (adviserinfo.sec.gov) built only from the target's stored CRD, and that link shows in every state, including when our 2024 archive holds no usable schedule. The redundant "Check stated fees" button is gone, and so is the "No applicable schedule" pill, which said only that the archived 2024 brochure had no schedule and could be read as a claim about today's fees.

              Sending a data request, your way or ours

              Two ways out of the draft, and neither one sends anything without a human clicking.

              • "Send from my email" is a split button offering Gmail, Outlook (Microsoft 365), Outlook.com (personal), your default mail app, and Copy email. Each opens a real compose window with the request pre-filled (Copy email puts the whole draft on your clipboard).
              • The choice you make is remembered, so the next request opens with "Open in Gmail" and a small arrow to change it. A draft too long for a compose URL is copied instead, with a note saying why.
              • "Send via ValuationMCP" opens a review of the exact To, Subject and Body, and nothing leaves until you click Send. It is signed "<your name> via ValuationMCP", replies go to you, and you are copied, so the target's answer reaches your inbox.
              • That email is always the standard request wording and never carries the valuation, the walk-away ceiling, a price or any figure; edits typed into the draft above still go out through your own mail client.
              • Every delivery is recorded against the request (which way it went and when, and for app sends whether it actually left) so the request shows "Sent via ValuationMCP on Oct 26, 2026", "Opened in Gmail on …" or "Email copied on …", including a failed or opted-out send.
              • App sends are capped at twenty per person per day and the same request cannot be re-sent within ten minutes; when a limit or a check cannot be confirmed, sending is refused and you are pointed at your own mail client.
              • The NDA gate is untouched: it is checked exactly as before, and an assistant creating a request through MCP or A2A still gets only the link and the draft. Sending exists solely behind the button in the app.

              Advertising measurement, with a disclosure and an off switch

              Four Google Ads conversion labels, the purchase one guarded by the payment provider's own numbers.

              • Successful account creation, an actual trial start, a contact/demo submission and a paid checkout each count once, on the production domains only.
              • A purchase event uses the amount, currency and transaction id the payment provider itself recorded, not a figure the page assumed, and refreshes or revisits of the success page do not recount it.
              • There is no consent banner. Instead, visitors in the EEA, United Kingdom or Switzerland get no advertising tags and no conversion events unless they choose otherwise, and a failed region lookup is treated the same way, fail closed.
              • Global Privacy Control and Do Not Track are honored, a refusal persists and is re-read before every send, and blocked events are never replayed later.
              • The privacy policy gained an Advertising measurement section saying exactly what Google Ads receives, that names, emails and practice figures are never sent, and how to opt out; the control sits on that page.

              Smaller changes

              • The public header on phones is a single sticky row with a top-sheet menu that marks the current page, and the sheet keeps keyboard focus and dismissal behavior.
              • The supported-servers directory gained a row of category chips with live counts that jump to each section.
              • The enterprise landing page stacks three use-case callouts (broker-dealers and enterprise RIAs, acquirers and aggregators, asset managers and custodians) and choosing one preselects the matching interest in the demo request, and the matching firm type for asset managers and TAMPs.
              • The Workspace, the Valuations page and the workbench now scope their practice and valuation queries to the signed-in account explicitly, so a stale or shared client can never read another account's rows.
              • The "Review before sending" window for a ValuationMCP-sent request no longer scrolls sideways: the secure link wraps inside the box, the body scrolls on its own, and Cancel and Send stay in view.

              v1.0.22

              A network you can run, plan changes you control, and the pricing page kept honestOctober 24, 2026

              The Network plan becomes a product: a network dashboard that rolls up members' latest saved valuations (outputs only, never raw inputs) email invites with explicit consent, and a report brand that co-brands exported PDFs. Plan changes are self-serve: an upgrade applies now with a prorated invoice, a downgrade waits for the renewal date, and a founding member keeps the $290 annual price for life. The pricing page's limits are now enforced where the work happens: practice and run quotas in the app, a view-only Free workbench past the limit, and Deal gates on the buy-side web endpoints. The home page gains the Form 5500 dataset and a fourth source card, and the primary nav finally works on phones. Owners gain a free outside-in read of their own firm from public Form ADV data, and buy targets show the fees stated in their Part 2A brochure. Engine 1.8.0 unchanged; server 1.24.0; eighteen tools: `outside_in_read` is new, so reconnect after publication to load it.

              The pricing page's promises, enforced

              Every limit the plans advertise is now checked where the work happens, not just at the assistant.

              • Saving practices and recording valuations count against the plan's quotas in the web app itself, through the same shared server paths the assistant uses; an over-limit write is refused with the upgrade block, never an error.
              • A Free account's web workbench goes read-only once the free full valuation is spent: the saved headline, range and coverage stay visible, while editing, re-running and recording are locked until an upgrade.
              • The buy-side web endpoints (data requests, deal-room rehearsal, buyer-hosted rooms and the diligence list) now require Deal, exactly like the buy lane itself.
              • Opening a seller-hosted shared room requires a paid plan.
              • Entitlement reads fail closed: if a plan cannot be confirmed, the restricted action is refused rather than quietly allowed.

              Network management

              One owner, invited members, and a dashboard of saved results.

              • A Network owner creates the network with its total advisor count, invites members by email-bound links and manages the member list, revoking an invite or removing a member, all under Settings.
              • A member explicitly consents before joining; leaving is self-serve on the member's side.
              • The dashboard rolls up each member's latest saved valuation outputs, headline, fair value, low, high, confidence and engine version. Raw inputs and source documents stay private to each member.
              • Membership grants Network access only while the owner's own Network entitlement is active; when it lapses, members revert to their own plans.

              Co-branded reports

              Display-only branding on what you hand to clients.

              • The network owner sets a report brand; exported valuation reports and deal briefs for members carry the co-brand line. Server-rendered chart branding comes later.

              Plan changes, self-serve

              Upgrades now, downgrades at renewal, and a founding price that never expires.

              • An upgrade applies immediately with a prorated invoice; a downgrade is scheduled for the renewal date, so the plan already paid for runs through its period.
              • Founding membership persists separately from the subscription: a founding member who cancels and later returns is still offered the $290 annual price for life.

              The outside-in ADV read. See what a buyer sees

              A free, unlimited read of your own firm built only from public Form ADV data, the buyer's math, before you connect anything.

              • An owner searches their firm by name or CRD in the Workspace, See what a buyer sees, and gets the same outside-in read buyers get on a target: filed assets and regulatory client count in, screening revenue derived from the published revenue-per-client study, and the buyer's walk-away ceiling with no sell premium.
              • Always a range plus a data-confidence score, never a single point value.
              • Free on every plan and unlimited: it never consumes the one free full valuation run and is never plan-gated.
              • Disclosures are explicit: which inputs are estimated, which methods cannot run without private data, and any engine-default multiple is labeled as such. A firm not covered by Form ADV falls back to a short manual form (assets, households, optional revenue) instead of an empty state.
              • Saving the read as a practice requires a verified firm claim, reusing the existing claim flow; an unclaimed lookup shows the read but cannot save it.
              • Below the read, an upgrade panel points to the full valuation. Connect your ledger and CRM to narrow this range, with the free run if unused, otherwise the upgrade.
              • New `outside_in_read` tool returns the same range, ceiling, confidence and disclosures, with an in-chat upgrade link for the full run. It is the eighteenth tool, so reconnect after publication to load it; server 1.23.0.
              • Funnel events record adv_read_viewed → firm_claimed → first_full_run → paid, tagged web or MCP, so the ad-to-upgrade path is measurable.

              Stated fees from the Part 2A brochure

              What a buy target's brochure says it charges, filed evidence beside the plan book, never an engine input.

              • A buy target's card shows the fee schedule from its Form ADV Part 2A brochure via advmcp.ai, opening with the source the way advmcp.ai does: As stated in the firm's Part 2A brochure, dated Mar 31, 2024, with its age and a View the brochure on IAPD link, then any newer-brochure warning and the caveat that a stated schedule is not what clients actually pay.
              • Each fee row keeps its verbatim excerpt and page number. Schedules are sorted conservatively into a usable tiered ladder or flat rate, a choice between several schedules (the buyer picks, recorded as its own provenance row), or not applicable with a one-line reason, fund fees, hourly or fixed only, negotiable only, incomplete or overlapping ladders.
              • Where a schedule applies, a non-blocking advisory compares the stated rate at the average relationship size (AUM ÷ clients) with the realized yield (revenue ÷ AUM). When the brochure doesn't say whether tiers blend, both blended and cliff rates are shown; every derived figure shows its formula and inputs.
              • Source is the SEC's 2024 brochure archive; current-year brochures are not included. Brochures are fetched only when you choose Check stated fees, and the schedule never counts toward coverage or confidence. Deal-tier only; downgraded Deal accounts keep read-only access.
              • Buy-lane saved runs snapshot the schedule, and their reports gain a Stated fees (Part 2A) section; older runs say no stated schedule was recorded. Sell reports are unchanged. `upsert_practice` and `list_practices` carry the schedule additively; engine 1.8.0 unchanged; server 1.24.0; still eighteen tools.

              Smaller changes

              • The footer names the HnyComb network, WealthMCP · ValuationMCP · ADVmcp · 5500mcp, with links that match 5500mcp.ai's hover and color states; external links open in a new tab.
              • The home page's proof strip adds the Form 5500 dataset (164,657 plans · 390,591 Form 5500 filings, refreshed September 2026) beside the ADV count, and the source-card row grew to four with a Form 5500 card that says what the filings are used for: finding large retirement plans that name a target adviser on Schedule C and what they filed paying it, a filed floor against claimed plan revenue, Deal-tier only and never an engine input. The cards now read Sica Fletcher, ADV, 5500, then Echelon, and the comparison card is titled the InvestmentNews 2025 benchmarking study.
              • The home page hero becomes What's your practice worth? Ask your AI assistant. (owner-facing copy now says assistant, never agent) with the lane line Selling? An asking price. Buying? Your walk-away ceiling. and a trust line under the CTAs: no standing connections to your systems; every pull is a one-time request in a conversation you can see.
              • The home page CTAs read Deal Room and A2A Compatible.
              • Private prep deal rooms replay the saved negotiation as an animated conversation between the two agents, with pause, show-all and replay, quicker playback, and reduced motion showing the final state.
              • The trust band no longer reads blue in light mode.
              • The primary nav works on phones: a hamburger menu with the current page marked, a smooth open and close, reduced motion respected, and the menu button beside Sign in.
              • Networks under Settings are visible only to Network accounts, and the connect page shows the A2A enterprise connection details, endpoint and Agent Card, only to signed-in Network accounts.

              v1.0.21

              Plans you can buy, an NDA before figures move, and a site that explains itselfOctober 23, 2026

              Pricing is public and self-serve: four plans, an in-app checkout that follows your light or dark setting, and a new `account_status` tool so your assistant can tell you what plan you are on and how many runs are left. Data requests, inbox items and agent answers can now sit behind a mutual NDA between the two firms: the target uploads an executed document (the counterparty acknowledges it) or waives the gate, and nothing seller-side prefills or submits until it is active. The web workbench saves to the practice and records a valuation through the same server paths the assistant uses, and new accounts open empty. Buy targets gain a Form 5500 retirement-plan book filed through 5500mcp.ai: Deal-tier evidence, not engine input. The public site was rebuilt: a new home page, public Skills, Developers and FAQ pages, and one header across all of them. Paid plans now flag practices due for re-value. Engine 1.8.0 unchanged; server 1.22.0; seventeen tools: `account_status` is new, so reconnecting loads it.

              Quarterly re-value nudge

              A timing reminder, never a new valuation.

              • Owner, Deal and Network flag active practices in either lane when their most recent saved, non-archived run is more than 90 days old. No saved run means no re-value flag; Free and archived practices are excluded.
              • A session-dismissible workbench notice offers a saved re-run and the What changed since last time? skill; practice lists and Targets carry a quiet badge. A fresh saved run clears the flag.
              • Additive `revalue` fields on `list_practices` and `revalue_due` on `account_status` let the assistant offer the same next step once without estimating or restating figures. Server 1.22.0; no new tool and no engine change. Skill guidance is staged only, not applied.
              • Due-ness is read from caller-owned saved dates; no cron or email sender. Email reminders are coming, blocked on an email domain.

              Plans, pricing and checkout

              Four plans, one source of truth, and a payment provider that is the only thing allowed to change your plan.

              • `/pricing` lists Free ($0), Owner ($39/mo · $390/yr), Deal ($149/mo · $1,490/yr) and Network (advisor-tier pricing based on the total number of advisors in the network, arranged with us). A monthly/annual toggle prices the self-serve plans; Network is not self-serve and its button opens the contact form.
              • Founding member: Owner billed annually at $290/yr for the first 100 members. The price is picked automatically while slots remain, and the page shows how many are left.
              • Checkout opens inside the app, and the payment screen follows your theme (dark background and violet button in dark mode, light background and violet button in light mode) with ValuationMCP as the business name. `/app/billing` shows the plan, the renewal date and what you have used, and "Manage billing" opens the provider's portal to change or cancel.
              • Only the payment provider's signed notification writes a plan, and each notification is applied once: a replayed event changes nothing, a forged one is refused, and a plan set by hand is never overwritten.
              • What a plan unlocks is enforced where the work happens. Free takes one practice and one full valuation, after which reruns return the headline, the range and the coverage only; the buy lane, data requests and deal-room rehearsal need Deal. A blocked call answers with an upgrade block (what you get, the price, the link) never an error, and a locked skill returns its title and blurb but not its instructions.
              • New `account_status` tool: plan, renewal date, runs and practices used, allowed skill themes, and a billing link, so an assistant can answer "what plan am I on" or "how many runs do I have left". It is the seventeenth tool, so reconnect after publication to load it.

              The NDA gate

              One agreement per firm pair, checked by one server-side path shared by the web, MCP and A2A.

              • Every new data request is created gated: `create_data_request` takes NDA options, its dry run returns the NDA preview alongside the request preview, and the returned email draft gains one line about the NDA step. A pair that already has an active agreement skips straight to the fields; existing open requests stay ungated.
              • `get_data_request` and `list_incoming_requests` return `nda.status` (awaiting signatures, awaiting acknowledgement, active, waived) before any field data. `submit_data_request`, including its dry run, refuses with `nda_required` and the signing/upload URL until the agreement is active or waived; no field prefill from the seller's practice before that either.
              • The request's expiry clock starts when the NDA becomes active or is waived, not when the request is created, so the window is not consumed while signatures are pending.

              Upload, acknowledge, or waive

              The executed document lives in private, write-once storage; only people open the gate.

              • Either party can upload a fully executed NDA PDF, but the gate opens only after the counterparty acknowledges it: "I confirm this is the executed NDA between our firms", and the acknowledgement is recorded. Waiving is target-only and recorded (who, when); the requester can never skip the gate.
              • Uploading, acknowledging and waiving require sign-in plus a verified CRD claim, offered inline on the request page so a recipient is never stranded.
              • Once executed, both parties download the document through a short-lived signed link. Agreements are listed in a new Agreements section in the Workspace, and the request page, the Targets row, the firm's inbox item and shared deal rooms all show the pair's agreement status. Deal rooms display status only; the gate does not block rooms yet.
              • No assistant tool signs. The agent can prepare the filled template and read the status; signing, uploading and acknowledging happen on authenticated pages by people. The requester's note stays untrusted information.

              A draft template, awaiting counsel

              A starting mutual NDA stored as a versioned template, marked DRAFT, pending counsel review and kept out of production use until the owner approves a version.

              • The template covers mutual confidentiality; a purpose limited to evaluating a potential transaction between the two RIAs; client information handled consistent with the Reg S-P business-transfer exception; the distinction between aggregate practice figures and client-level data; an optional non-solicit of clients and staff; return or destruction of information; a two-year default term; and governing law as a merge field across US states and Canadian provinces.
              • The assistant never writes or edits NDA text; it fills merge fields only, legal names, CRDs, addresses, purpose, effective date, term, governing law.
              • Built-in e-signing with a certificate PDF, email notices, and DocuSign or Dropbox Sign come in a later phase. The `nda` skill is drafted and staged, not applied yet.

              The workbench saves what you enter

              One save path, shared with the assistant.

              • Save to practice and Record valuation in the web workbench go through the same server paths the assistant's `upsert_practice` and `compute_valuation` use, so what the web saves and what an agent saves cannot drift apart.
              • The workbench takes the same raw book-quality facts the connector accepts, twelve inputs across client demographics and succession readiness, and derives identical scores server-side with the same fixed anchors.
              • A new account opens an empty workbench: no example inputs, no example valuation, nothing until a practice is selected.
              • A practice with assets but no EBITDA on file no longer opens with a fake "Unsaved changes" prompt or writes an estimated survey EBITDA on save; the placeholder appears only after an explicit choice, and a save with a sourced default keeps its citation.

              Form 5500 plan book, from 5500mcp.ai

              Filed retirement-plan evidence beside the buy-lane workbench, figures the plan administrator itself filed, never an estimate.

              • A buy target gains a retirement-plan book pulled from DOL Form 5500 filings through 5500mcp.ai: the buyer confirms the adviser's EIN (offered from the ADV filing), and the book lists the plans, participants, assets, provider compensation and the target's share of it, as filed. The whole feature is Deal-tier, gated like the buy lane itself.
              • Where a filing leaves the adviser's EIN blank on Schedule C, a plan joins the book on an exact match to the ADV legal or business name, and such rows are labeled as name matches.
              • The card sits on the workbench under "Request data from target" and on each Targets row: the plan table, aggregates, a "filed floor" line for what the filings prove about provider revenue, and fixed caveats. Where a figure is not on the form it reads "Not filed on this form", never zero.
              • `upsert_practice` gains `adviser_ein`, `refresh_plan_filings` and `plan_refs`; `list_practices` returns a plan-book summary for buy-lane practices. Manual entry by sponsor EIN and plan number works today, and provider search switches itself on when 5500mcp adds it; the capability is checked from the tool's own input schema.
              • Each saved buy run snapshots the plan book and the plan-book text is appended after the verbatim report, so report text itself is unchanged. Engine 1.8.0 untouched: the plan book is filed evidence, not an engine input.
              • The book lives in one provenance row, DOL Form 5500 via 5500mcp.ai, with an empty supplied-fields list, so it never counts as coverage. A new optional data-request field, Retirement-plan revenue, is display only. A confirmed EIN stays editable; every change writes a new provenance row.
              • Fixing a plan book from an assistant no longer fails when nothing else on the practice changed.

              A public site that explains the product

              The home page was rebuilt, the old one archived rather than deleted, and three public pages added.

              • The new home page walks from a worked example through the sell and buy lanes, how it works, the sources, the skills and the developer path, quoting the live figures it shows (filings indexed, supported servers, tools) read from the same data the app uses.
              • The previous home page is archived at `/home-legacy` as a reference copy: noindex, and out of the navigation, the footer and the sitemap.
              • `/skills` and `/developers` are public: the whole skill catalog with its one-page artifacts, and the MCP surface with the tool list and connection notes.
              • `/faq` holds every question in five URL-addressable tabs (Getting started, Selling a practice, Buying a practice, Enterprise and succession, Pricing) as accordions, and the full set is published as structured data so search engines can show the answers. The home page carries the first six and links to the rest.
              • One header now sits on every public page, on the logo's own line: How it works, Sell or buy, Sources, FAQ, Developers, then Pricing, Contact and a single Sign in button. Two buttons that opened the same page was one too many.
              • The home page's opening conversation plays once when it comes into view, pauses while you are looking elsewhere and resumes where it stopped; with reduced motion on, or without JavaScript, the finished card is shown complete.
              • Type is now Schibsted Grotesk for every word and Martian Mono for every figure (in the app, the charts, the exported PDFs and the one-page artifacts) so a printed brief reads like the screen. The largest dollar headlines were sized down.

              Smaller changes

              • Buy lane: the Request data from target block sits with the valuation in the workbench, and the same request flow replaces the old Gather new data steps for a buy target in the workspace. Selling is unchanged.
              • Deal rooms can be archived without deleting anything: an archived room disappears from every list and its invite links read as invalid, while every row stays in the database.
              • On the free example page the Download PDF button sits beside the firm's name instead of under the brief heading.
              • Browser requests for the Agent Card URL now land on the Agent-to-Agent page; agents still receive JSON, and the card itself is unchanged.
              • The home page status pill moved from Alpha to Beta.
              • The request page's connect-your-assistant section is inline and expandable: the prompt to use and what signing in does, with no detour to another page.
              • A reply from a target is visible where you asked for it: the Workspace request button carries a "1 to review" badge and the request panel opens with a "Response received" banner. The figures stay unsaved until you accept them.
              • The answer form takes decimals again; a value such as 58.4 was being refused by the browser, while counts like households stay whole numbers.
              • The Agent-to-Agent page's "Find your assistant" button was removed: the server catalog lists data sources, not assistants.
              • Contact is a real form now: `/contact` takes a name, email, firm, topic and message and saves them for the team, reachable from the header, the footer and the Network plan. It holds five messages an hour per visitor and three hundred a day site-wide, with a hidden field that catches bots. No email alerts yet, messages wait to be read.
              • The footer's connector link reads "Supported MCP Servers" without the number, and the Agent-to-Agent and Deal Room links moved out of the footer.

              v1.0.20

              18 new MCP servers, a Public filings section, and live server countsOctober 20, 2026

              The supported-servers directory grows from 38 to 56 listings (52 distinct servers across 8 source categories), after a full review of the WealthMCP directory for servers that can supply a valuation input. Every new server has its own detail page and appears in the Connect wizard. Site content only: no tool was added or removed, no reconnect is needed, engine 1.8.0 and server 1.18.0 are unchanged and no computed figure moved.

              New servers

              Each new listing can supply at least one valuation input.

              • Custody & clearing: Interactive Brokers MCP Server (community), Fidelity MCP Server (claude-fidelity, community).
              • Portfolio accounting & wealth platforms: iCapital Connector for Claude, Milemarker Firm MCP, Wealth.com (Wealth MCP), eMoney MCP Server (community).
              • CRM: Zeplyn MCP Server, GoHighLevel (LeadConnector).
              • Fee billing: Pocket Invoice MCP.
              • Accounting ledger: Narmi / Grasshopper Bank, Lili.
              • Firm intelligence & registries: Fintel Discovery.
              • Public filings (new section): 5500mcp, PlanTrust, PlanProvider.Pro, EdgarTools, SEC EDGAR MCP Server, SEC EDGAR (secedgar).

              Other changes

              A new section, counts that can't drift, and a clear scope.

              • New "Public filings" section on /mcp-servers for Form 5500 and SEC EDGAR sources. 13F holdings value is labeled as a proxy for AUM, not regulatory AUM.
              • Live supported-server count on the home page, /mcp-servers, every server detail page, the Connect wizard, the connect-your-assistant card and the footer, derived from the catalog so it can't drift. The headline counts distinct servers; per-section counts include servers listed in more than one section.
              • Scope note: the catalog lists US-market servers; ledgers and platforms serving only non-US markets were left out.

              v1.0.19

              MCP-native, A2A-compatibleOctober 19, 2026

              Enterprise assistants can now use the same private information-request flow through A2A, alongside MCP. The buyer approves the request and the firm owner approves the answer; neither side’s valuation is shared.

              Same middle, another way to connect

              ValuationMCP stays in the middle. A different connection does not mean different permissions.

              • An Agent Card introduces the request broker and its two skills: request data from a target, and answer a data request.
              • Assistants sign in with OAuth and work within the same account and firm-claim boundaries as MCP. A preview saves nothing; sharing still needs explicit human approval.
              • Requests move from waiting for the firm to buyer review, then completion or cancellation. Buyers can cancel an open request; firms decline in the app.
              • The buyer sees requested answers and any shared source lines, never the firm’s valuation. The firm never sees the buyer’s walk-away ceiling.

              v1.0.18

              Requests straight to your firm's inboxOctober 18, 2026

              When both sides use ValuationMCP, buyers can deliver requests here without email. Firms verify their CRD and choose to receive requests. Engine 1.8.0 unchanged; server 1.17.0; sixteen tools.

              Direct delivery, by choice

              Your valuations stay private. Receiving requests does not mean your firm is for sale.

              • Settings → Your firm matches a confirmed account email to the website on Form ADV, or sends the claim for admin review. Receiving requests starts off and must be switched on.
              • Buyers can choose a private link or direct inbox delivery when available. Firms that have not opted in remain invisible to buyers.
              • The firm sees the sender's stated name and firm, plus an account-domain or reviewed-firm identity line, never the buyer's full email.
              • Requests for your firm shows incoming requests. Answer with the same editable form and review step, decline without a reason, or block the sender. Open inbox requests are limited to three per CRD and one per buyer–CRD pair.
              • A firm's assistant can list incoming requests, read by request id and preview an answer. Only explicit owner approval sends it. Declining stays in the app. Accepted answers follow the existing saved-run process.
              • Reconnect your assistant after publication to load the new tool.

              v1.0.17

              Agent to agent: buyers ask, firms answer from their own assistantOctober 18, 2026

              A buyer's assistant can prepare a private request and draft the email for the buyer to send. The firm's assistant can read a request, gather figures from its own records and show a preview. Sharing happens once, only after the owner approves. Engine 1.8.0 unchanged; server 1.16.0; fifteen tools.

              Owner-approved answers

              Sharing stays voluntary; the firm's own valuation stays private.

              • A buyer's connected assistant can preview a request for one of the buyer's own targets, show the draft email, and create it only after the buyer approves. The link is shown once, and the email is left as a draft for the buyer to send.
              • A connected assistant can read a received request and prefill only the requested fields from the firm's own practice. The requester's note is information, not instructions.
              • Before sending, the assistant shows the owner each figure and the source line the buyer will see. The owner can remove any field or decline; only explicit approval sends the one-time answer.
              • Buyers see when the firm's AI assistant submitted an answer. Figures reported from another system are labeled as not checked against a ValuationMCP record, and remain leads for diligence like typed figures.
              • Accepted answers follow the same review and saved-run process as the private web form. Reconnect your assistant to load the three new tools.

              v1.0.16

              Buyer-hosted deal roomsOctober 18, 2026

              A buyer can open a shared room from a saved buy-side run and invite the target. The walk-away ceiling is never shared; the room carries the target data request. Engine 1.8.0; server 1.15.0; twelve tools.

              Buyer-hosted deal rooms

              The mirror of the seller-hosted room, on the same machinery, with the walk-away ceiling never shareable.

              • A buyer can now open a shared room from a saved buy-side run and invite the target. It mirrors the seller-hosted room and runs on the same machinery: invite by email or link, tick what is shared, threaded comments with Jev checking every comment, a neutral AI agenda, guest removal and link control.
              • The walk-away ceiling is the buyer's private anchor: it is never shared, never shown to the guest, and no tick exists for it. The default shared set is the target's own filed facts (AUM, regulatory clients, filing date, CRD) and the inputs diligence still needs; optional ticks add coverage, accepted target-provided figures and the published market context, each with a one-line warning.
              • The room carries the existing target data request: an open request is reused, the submission is immutable, accepted figures keep target-provided provenance and re-run the ceiling, and the guest sees only their own submission status, never the resulting valuation.
              • A private buyer coach mirrors the seller coach: it matches the target's comments to the buyer's private rehearsal memo, visible only to the host. Jev blocks a buyer comment that would leak the ceiling or any unshared figure, and tells the buyer why privately.
              • list_practices shows each practice's deal room (status, side, created, guest joined) without links or tokens. Engine 1.8.0 unchanged; server 1.15.0; twelve tools.

              v1.0.15

              How the screen held upOctober 18, 2026

              When a target's own figures arrive, the buy report compares the screening ceiling with the diligence ceiling input by input. Engine 1.8.0 (output only; no existing figure changes); server 1.14.0; twelve tools.

              Screening-to-diligence comparison

              • A buy target with an earlier screening run and later target-provided figures now gets a "How the screen held up" section in the report and workbench, and a Screen vs diligence change on the Targets tab linking to the run.
              • The engine rebases the screening run on the current engine, then swaps inputs one at a time in a fixed order (revenue, EBITDA, EBOC, cash flow, clients/AUM, growth, then each quality factor) and records each step's effect on the walk-away ceiling. What the single steps don't explain is shown as an interaction remainder, so the rows always add up to the total.
              • Each input reads held (within ±10% of the diligence value), moved (outside ±10%) or new (a default or missing before). A not-screened baseline reads "No screening ceiling existed" with the first ceiling and the inputs that arrived.
              • compute_valuation returns screeningDelta, stored with the saved run and readable through list_practices history; Explain a valuation change is starred when it exists. Sell lane unchanged.

              v1.0.14

              Ask the target for its own figuresOctober 18, 2026

              Buy-side targets can now receive a private data-request link; accepted figures save as target-provided and re-run the ceiling. Engine 1.7.0; server 1.13.0; twelve tools.

              Target data request link

              • A buyer can send a private link asking a target for its own figures: TTM revenue, EBITDA, EBOC, free cash flow, households, AUM, growth, retention, recurring share, concentration, client age, decumulator share and the succession facts. The link expires in 14 days by default (30 at most), can be revoked, and can be extended once. One open request per target.
              • The target's page needs no account: it shows who is asking and why, every field is optional, consent is required, and nothing about the buyer's valuation, thesis or pipeline is ever shown. One submission per link; expired or revoked links simply read as unavailable.
              • The buyer reviews each submitted figure next to what is already on file, with differences flagged, and accepts all, per field, or rejects. Accepted figures save as target-provided provenance and trigger one new saved run labeled "After target data request", a not-screened firm becomes a real walk-away ceiling once its revenue is accepted.
              • list_practices shows each target's request status and timing (never the token), and compute_valuation's What-next notes when target data is waiting for review. Engine unchanged at 1.7.0; server 1.13.0; still twelve tools.
              • Server 1.13.1 hardening: submitted target figures are locked once received, so anything labeled target-provided is exactly what the target sent. Buyers can no longer edit requests or submissions directly; revoke, extend (once, never past 30 days from creation) and accept/reject run server-side after an ownership check, and a closed or expired request can never be reopened.

              v1.0.12

              Non-retail targets are not screenedOctober 18, 2026

              Excluded buy targets need target financials before any estimate or rehearsal. Earlier saved results are preserved. Engine 1.7.0; server 1.12.0; twelve tools.

              Non-retail targets are not screened

              • New excluded buy runs with no supplied revenue produce no valuation, range, method table or sensitivity. They show Not screened, target financials required, filed evidence, coverage, diligence requests, peers and market facts.
              • Target revenue leads What diligence replaces. Buyer diligence and data-confidence triage are starred; rehearsal is blocked before any credit spend or memo save until target revenue is on file.
              • Earlier saved excluded valuations and memos remain available as saved. Supplied-revenue buy runs and the sell lane are unchanged.
              • Engine 1.7.0; server 1.12.0; still twelve tools. Append-only buyer skill guidance is staged, not applied.
              • Server 1.12.1: a saved not-screened run advertises and renders only the peer-position and market-tiles charts (and a pack of just those two); the estimate charts return Not found.

              v1.0.11

              A flagged firm no longer reads as a ceiling, and screening no longer grades itselfOctober 18, 2026

              A buy-side run whose filing falls outside the retail practices the published medians describe now drops the Client Value method when it rests on the retail median, and reads "Remaining-method estimate, not a screening ceiling" everywhere instead of "walk-away ceiling". Screened runs no longer compare their derived revenue per client against the medians it was derived from, and the folded advmcp.ai peer cohort in get_benchmarks answers again. Engine 1.6.1; server 1.11.1; still twelve tools.

              A flagged firm no longer reads as a ceiling

              • When a firm's filing falls outside the retail practices the published medians describe, the Client Value method is now excluded too when its value per client rests on the retail industry median: the median describes exactly the firms the check ruled out. A firm-derived or supplied value per client is unaffected.
              • Those runs no longer say "walk-away ceiling" anywhere: report, workbench, Targets, charts, PDF or rehearsal memo. They read "Remaining-method estimate, not a screening ceiling", with one sentence that target-provided revenue is required before any ceiling. Runs saved before this change read back as saved.
              • The confidence line is capitalized: "Not a meaningful screen: target financials required".

              Screening no longer grades itself

              • On a screened run the revenue-per-client note in "What's helping / holding it back" and the plausibility line in "Before you rely on this" are suppressed. That revenue was itself derived from the same study medians, so the comparison graded the derivation against its own source.
              • The derived value-per-client basis on screened runs now says what it is: log-AUM interpolation between the study's tier medians, not a flat tier median.
              • The folded advmcp.ai peer cohort in `get_benchmarks` answers again: the request sent a parameter the endpoint rejects, and it is no longer sent.

              Smaller fixes

              • The home page lists all twelve tools, including `server_info`, and mentions the CRD filing fill and the Targets pipeline.
              • `list_practices` returns `thesis`, `stage`, `stage_updated_at` and `filing_as_of`, so an assistant can list targets by stage.
              • Refreshing a filing updates the existing filing source instead of logging a duplicate; exact duplicates are removed, keeping the newest.
              • The buy-lane What-next menu offers Deal Room Rehearsal: "Rehearse the purchase of <name>, what will the seller argue?"

              Skills and compatibility

              • Still twelve tools; no new or changed inputs. Server 1.11.1, engine 1.6.1. Sell-side figures and buy-side runs with supplied revenue are unchanged.
              • Skills updated: Buy-Side Ceiling 2.2.1 and Find and Screen Acquisition Targets 1.5.1 now describe the CRD filing fill and the non-retail check.

              v1.0.10

              Screening ceilings you can trust, Form ADV facts from a CRD, and a Targets pipelineOctober 17, 2026

              A buy-side target valued from public data now gets a screening revenue the engine derives and cites, smooth across firm sizes instead of jumping at tier boundaries, with a table showing how the ceiling moves if that assumption is wrong. Give a buy-side target its CRD and the server fills its filed assets and client count from the Form ADV filing. The app gets a Buying / Selling switch and a Targets view that tracks each target by thesis and stage. Firms whose filings don't look like a retail practice are flagged instead of screened. Engine 1.6.0; server 1.11.0; still twelve tools.

              Screening revenue, owned by the engine

              • When a buy-side target has filed assets and clients but no revenue, the engine derives a screening revenue: filed clients × revenue per client from the InvestmentNews 2025 study (2024 data), blended smoothly across asset sizes. It's labeled benchmark-derived, cited, and listed under "What diligence replaces". Revenue you supply always wins.
              • The blend removes the jump at tier boundaries: two otherwise similar firms at $240M and $260M of assets now land about 4.5% apart instead of 47%. The pin points for the two open-ended tiers ($177M and $1.41B) are ValuationMCP's choice and are labeled that way; the dip above about $707M comes from the study's own over-$1B median.
              • A new assumption table, computed by the engine, shows the ceiling with screening revenue 20% lower and higher, against the low, middle and high of the published index.
              • Screening runs say "a screening ceiling: filed facts plus cited medians", so low coverage reads as what it is.
              • Firms with less than 50% of filed assets held for individuals and high-net-worth individuals are flagged as "Screening not meaningful", using the documented SEC IAPD retail-focused subset rule. When filed client mix is unavailable or incomplete, assets per regulatory client above $24,503,354.11 trigger the same flag | ValuationMCP's chosen conservative cutoff of ten times the August 12, 2026 IAPD retail p75 ($2,450,335.41), not a published p95. No screening revenue is derived; remaining methods still compute, confidence says "not a meaningful screen: target financials required", and revenue leads "What diligence replaces". Supplied revenue bypasses the check.
              • Engine 1.5.0 introduced screening revenue; 1.6.0 adds the non-retail check. Sell-side figures and buy-side runs with supplied revenue are unchanged.

              Form ADV facts from a CRD

              • Create a buy-side target with its SEC CRD and the server fetches the firm's Form ADV from advmcp.ai, fills its filed assets and regulatory client count with the filing date, and logs the filing as the source.
              • Figures you supply win; where they differ from the filing, the report shows both and uses yours. If the filing can't be fetched, the target still saves with a note. `refresh_filing` re-fetches.
              • Sell-side practices never trigger a lookup, and nothing beyond the filed facts is inferred.

              Buying, selling and a Targets pipeline

              • The new-practice form asks whether you're selling a practice you own or buying a firm you might acquire; the workbench gets a Selling / Buying switch and a lane badge on every practice.
              • Recorded valuations has a Targets tab: buy-side targets grouped by thesis, each with its filed facts and filing date, latest walk-away ceiling with range and coverage, and a stage, screened, contacted, NDA, LOI, diligence, closed or passed. Stages are your own tracking labels, never a signal that a firm is for sale.
              • `upsert_practice` accepts `thesis` and `stage` for buy-side targets, so an assistant's screen lands in the Targets tab.

              Skills and compatibility

              • Still twelve tools. New optional inputs: `refresh_filing`, `thesis`, `stage`. An assistant connected before this release should reconnect once to see them.
              • Skills updated: Buy-Side Ceiling 2.2.1, Find and Screen Acquisition Targets 1.5.1, and Deal Room Rehearsal 1.2.2. They rely on the server's screening revenue and CRD fill instead of asking the assistant to compute or look anything up.

              v1.0.9

              The buy side: walk-away ceilings, buyer prep in the Deal Room, and archive instead of deleteOctober 17, 2026

              ValuationMCP now values both sides of a deal. A practice is created in a lane (sell for a practice you own, buy for a firm you might acquire) and the engine owns the difference: a sell-side run headlines an asking price with the usual listing premium, a buy-side run headlines a walk-away ceiling at fair value, priced off the published index low for the firm's size and labeled that way. The Deal Room follows the lane: a rehearsal on a sell-side valuation prepares the seller as before, and a rehearsal on a buy-side valuation now prepares the buyer, the seller's case for a higher price, the buyer's answers from the figures and the diligence gaps, the terms each side will push. Deleting a practice or a run now archives it, so nothing is lost to a misclick. Engine 1.4.0; sell-side figures are unchanged. Server 1.9.1; still twelve tools.

              Buy lane: a walk-away ceiling the engine owns

              Buy-side behavior used to depend on the assistant suppressing parts of a sell-side result. It now lives in the engine and the tools.

              • `upsert_practice` takes `lane` (`sell` or `buy`, set when the practice is created and fixed after) and an optional SEC `crd`. Existing practices are sell-lane.
              • A buy-lane run headlines a **Walk-away ceiling** equal to the fair estimate, with no premium. A sell-lane run keeps its asking price. Both come from the same seven methods.
              • On the buy side, revenue and EBITDA multiples you don't supply default to the published index low for the firm's size, labeled as the ceiling convention rather than as figures you supplied. Multiples you volunteer still win, and the sensitivity strip still spans the full published range.
              • The buy-lane report says "What diligence replaces" (the inputs that rest on medians, placeholders or gaps) and "The market you're buying into". The words "asking price" never appear in a buy-side result.
              • `record_data_source` accepts `public_filing` (reported as filed) and `target_provided` (reported as target-provided), so filed facts and a target's own financials are never mistaken for owner-stated figures.
              • `list_practices` can filter by lane, and the What-next menu after a buy-lane run offers the buyer workflows.
              • Engine 1.4.0. Sell-side reports and charts are unchanged apart from the engine-version stamp.

              Deal Room: both seats

              The rehearsal reads the saved run's lane and prepares whichever side you are on.

              • Seller prep is unchanged: the buyer's arguments, your answers, the terms each side will push, and a prep checklist.
              • Buyer prep is new: the seller's case for a higher price, your answers grounded in the run's figures and its diligence gaps, the terms each side will push, and a checklist of what to verify before raising your number.
              • Same rules on both seats: the model never states a number, every argument cites figures from the saved run, and Jev drops anything those figures don't support. A walk-away ceiling is never framed as an offer, and nothing in a memo suggests a firm is for sale.
              • Memos are labeled by seat in the app, in `list_deal_room_memos`, and in the PDF, buyer memos export marked "Private · Buyer only". `run_deal_room_rehearsal` takes the same inputs as before; the seat comes from the run.
              • The shared room stays seller-hosted with the buyer as an invited guest. A buyer's prep memo never appears in a shared room or in the seller's coaching.
              • The Deal Room page and the home page now describe both seats: seller prep, buyer prep, and the shared room.

              Archive instead of delete

              Buyers create a lot of target records. A delete is no longer permanent by accident.

              • Deleting a practice in the workspace, a run on Recorded valuations, or a practice in admin now archives it. Archiving a practice archives its runs; restoring it brings them back together.
              • An Archived view in the workspace lists everything archived, with Restore. Permanent delete is available only there, only for archived practices, and only after you type the practice name.
              • Archived practices are hidden from every picker and from `list_practices`. The valuation tools answer "That practice is archived. Restore it in the workspace first."
              • A shared room built on an archived valuation shows its members "This valuation has been archived by its owner" instead of an error.

              Skills and compatibility

              The tool surface did not change, so nothing breaks for a connected assistant.

              • Still twelve tools. `lane` and `crd` are new optional inputs and the new response fields are additive; an assistant connected before this release should reconnect once to see the new inputs. `server_info` reports server 1.9.1 and engine 1.4.0.
              • Skills updated: Buy-Side Ceiling 2.0.0, Find and Screen Acquisition Targets 1.3.0, Presentation Contract 2.5.0 and Deal Room Rehearsal 1.2.0. They now rely on the server's buy-lane output instead of asking the assistant to override sell-side defaults. Skills are fetched fresh on every call, so no reconnect is needed for them.
              • The workbench PDF now captures every workbench section, chart and input.

              v1.0.8

              The Deal Room: rehearse the sale, privately or with a buyer in the room, in the app or from a connected assistantOctober 15–16, 2026

              The Deal Room takes a saved valuation and argues with it. In sell-side prep it writes the price spread, the points a skeptical buyer will press, the owner's answer to each, the terms each side will push, and a prep checklist. Every figure copied from the saved run, nothing stated by the model. The rehearsal can also be handed over as a private prep brief, one page you reopen before each buyer conversation, built from that same saved run, and either brief, the private memo or the free example, downloads as a branded PDF drawn from the figures already on screen. In shared mode the seller owns a room, picks which figures an invited buyer can see (the range and data coverage always visible), and the two sides work a neutral agenda with threaded discussion. Jev runs the judgment calls throughout: it picks the right firm when a name matches several, checks each debate argument against the figures it cites, checks every shared-room comment before it posts for a figure that isn't shared or one the seller kept private, and links each buyer comment to the matching answer in the seller's private prep memo. A rehearsal run from a connected assistant goes through the same server-side generator as the app: two new tools, run_deal_room_rehearsal and list_deal_room_memos, and the Deal room rehearsal skill moves to 2.0.0, so the assistant presents the server's memo verbatim instead of writing its own; the prep artifact stays one static page, with the PDF living on the saved memo it links to. A free example runs the same debate on any registered adviser from public filings, with no account. Also in this release: engine 1.3.2, so a quality factor nobody supplied reads as neutral instead of inheriting a sample score, a figure change, and a figure filled from a published median now reports as benchmark-derived with its citation. Server 1.8.0, the tool list grew from ten to twelve, so a client connected earlier must reconnect once; the valuation engine stays at 1.3.2.

              New: the Deal Room, rehearse the sale before you are in one

              A saved valuation is the starting point, not the end of the work. The Deal Room takes that run and argues with it: what a buyer will say to bring the price down, and what the owner's best answer to each point is.

              • Run it from any saved valuation. The memo comes back in five parts: the price spread (where a buyer is likely to open, the fair-value estimate, the asking price), the arguments a skeptical buyer will press, the seller's counter to each one, the deal terms each side will push (structure, earn-out, client retention, transition, non-compete, described in words) and a short checklist for before the first meeting.
              • Every figure comes from the saved valuation, copied field by field, including the engine version that produced it. The model writes the arguments and names the figures it leans on; each argument carries a tag showing the figure and its source. Any line where the model tries to state a number is removed before it is shown, and the memo says how many lines were removed. Two memos can never mix facts from two different runs.
              • Memos are saved against the valuation they were written from and listed newest first. Earlier memos for the same valuation can be opened or deleted, and only your own memos can be deleted. A memo is never edited after it is saved, and starting a new one asks first, because it spends AI credits and only the arguments change; the figures do not.
              • Prep memos are visible only to whoever ran them. A buyer, when there is one, never sees them.
              • A guided workflow, "Deal room rehearsal", follows the same rules for a connected assistant, so the rehearsal can happen in the chat where the valuation was run.

              A shared room, when a buyer is in the conversation

              The Deal Room switches between two modes: private sell-side prep, and a shared room the seller owns and a buyer is invited into. This is the one place the app serves both sides, and the seller keeps control of what the other side can see.

              • You choose what the buyer sees, figure by figure. The value range and the data-coverage score are always shared, so a buyer never receives one figure without the range and the confidence behind it. Everything else (assets, client count, revenue, profit) is your choice, and can change at any time.
              • The shared figures come from your saved valuation, not from anything typed into the page. A figure the engine did not return cannot be added to a room.
              • A neutral agenda for the meeting is written from the figures you shared: the questions those figures raise and the terms to settle, arguing for neither side and revealing no negotiating strategy.
              • Discussion happens in place: each agenda item and each deal term has its own comment thread, plus a general thread. Every comment is labeled Seller or Buyer from the room's own record of who owns it, not from anything a person claims.
              • Invitations go by email address: a link only someone signed in with that exact address can use, or by a share link you can switch off at any time. You can remove a buyer, cancel an invitation that has not been used, and a buyer can leave.
              • Invites are not emailed automatically yet: you copy the link and send it, because the app has no sending address configured. Once a domain you own is connected, the app can send them for you.

              A free example on any registered adviser, with no account

              The Deal Room is explained and demonstrated in public, on a firm anyone can name, so the feature can be judged before anyone signs up.

              • Type a firm name or its CRD. A name searches the public filings and lists the matching firms with their CRD and city to pick from; a CRD starts immediately. Firm data comes from SEC Form ADV filings served by advmcp.ai, linked at every mention.
              • The same engine prices the firm, and the page shows the low end, fair value, asking price and high end. Filings carry assets and client counts but no revenue or profit, so those are filled from published industry medians, labeled as estimated from median, and every method they drive is marked provisional. Coverage counts only what the firm actually filed, so an estimate never looks better-sourced than it is.
              • Buyer and seller write their arguments at the same time, in two columns, as they are generated; the debate is visible rather than summarized afterwards. Lines that state a number are hidden before they appear, and each argument is tagged with the figure it relies on. When both sides finish, the brief assembles below.
              • It is a simulation, not a shared room with a real buyer, and the page says so: the full Deal Room runs on your own saved valuation with the financials and quality factors you supply. The two modes are illustrated side by side, and the homepage offers the example after the supported-servers button, marked New.
              • The demo spends AI credits without a sign-in, so it is capped: three runs per visitor per day, 150 site-wide, a seven-day cache so the same firm asked again costs nothing, and sixty firm searches per visitor per hour. If the limit check cannot run, the demo refuses rather than running unchecked.

              The demo's judgment calls are made by Jev, not a guess

              Two spots in the free example previously leaned on heuristics. Both now ask Jev (TypeSafe's System One decision model, called as typesafe/jev-latest through the Lovable AI Gateway) a typed question whose answer is a probability. It cannot write a figure, which keeps the demo's one rule intact.

              • Picking the right firm: when a name search matches several advisers, Jev decides which one the visitor most likely meant, and that firm moves to the top of the list with a Likely match badge. It only flags when it is confident, if it is unsure or none of the firms fits, the list appears plain, and typing a CRD skips the question entirely.
              • Keeping the debate honest: every argument on either side is now checked by Jev against the figures it cites before it is shown. An argument Jev judges unsupported by its own cited facts is removed rather than displayed, alongside the lines that state a number. The brief assembles from the surviving arguments only.
              • Both checks degrade safely: if Jev cannot be reached, the plain list and the arguments it would have filtered still appear; the feature narrows, it never blocks.
              • These judgments sit in the free demo, the full Deal Room's argument pass, and, new since the notes opened, the shared room below. The valuation engine, the MCP tools, and no computed figure moved.

              Jev works inside the shared room: a comment check and a private coach

              The same judgment model now runs in the shared room, on both sides of the table. It returns a probability and nothing else, so it can flag a comment or point at an existing answer. It can never state or suggest a figure.

              • Every comment is checked before it posts, the seller's and the buyer's alike. Jev asks two questions: does the draft state a financial figure that is not one of the figures shared in the room, and, when the seller has kept figures private, does it reveal or closely paraphrase one of those. A confident judgment (50% or more) shows the poster a plain-language warning with two choices: edit, or post anyway. The check nudges; it never blocks or silently removes anything.
              • A private prep coach for the seller: a "Match buyer comments to my prep" button in the seller's own view asks Jev which prepared answer in the private sell-side memo most directly responds to each recent buyer comment, and the matched answer appears under that comment labeled "Your prep (only you see this)". The room reports how many comments were checked and how many matched. It requires a prep memo to exist, the coach only links to answers already written, never writes advice, and is invisible to the buyer.
              • The shared agenda now runs the same grounded-argument check as the demo debate: an agenda argument Jev judges unsupported by the figures it cites is removed before the agenda is shown, and the count is reported with it.
              • Both features degrade safely: if Jev cannot be reached, comments post as normal and the coach says it is unavailable rather than guessing. No figure anywhere in the room comes from Jev; the room's numbers still come only from the saved valuation.

              The rehearsal can now be handed over as a prep brief

              Every artifact this app produces follows one contract: a page the user can reopen later, built only from figures a tool returned, with the caveats kept. The deal room rehearsal was the one workflow that stopped at the chat. It is now the tenth that can hand one over.

              • After the memo and the what-next menu, the assistant offers a private prep brief: "Deal room prep · <practice name>", a single page you reopen before each buyer conversation. A later rehearsal updates that same page instead of creating a second one, so its history and its link survive.
              • It carries what the memo produced and nothing else: the saved valuation it is built from (that run's date and engine version), the price spread with data coverage beside it, each buyer argument with the prepared answer next to it, the terms each side will push in words, and the prep checklist. It adds no new advice.
              • Every figure on it is copied from the one saved run named in the header. Arguments, answers and terms are words, none of them states a number, and the brief never proposes a price, a counteroffer, a walk-away figure or "fair terms".
              • It is private by construction: the page says it is a rehearsal, not something to send a buyer, and it carries nothing from a shared room's threads. The shared room gets no artifact at all; a live conversation with a guest is not a document to freeze.

              Either brief now downloads as a branded PDF

              A rehearsal you reread on screen is one thing; a page you can hand to an advisor or take into a meeting is another. Both briefs now download as a designed PDF, built from the same figures the page already shows.

              • "Download PDF" sits with the brief it belongs to: in the header of the free example's brief, and at the top of every private prep memo. One click and the file arrives: no print dialog, no reformatting by hand.
              • The file carries the brief as written: the price spread with data coverage printed directly beneath it, each buyer argument with the figures it leans on and the prepared answer beside it, the terms each side will push in a two-column table, and the prep checklist as tick boxes. An "About this brief" section and page numbers close it.
              • Which brief it is stays on the page itself: a dark header band carries the logo and a tag: "FREE EXAMPLE · PUBLIC FILINGS" on the free demo, "PRIVATE · SELLER ONLY" on a prep memo, so a file saved to a desktop can never be mistaken for the other.
              • The one rule holds inside the file: every figure is copied verbatim from the brief's own figures, and the PDF code never formats, rounds, or computes a number. Nothing appears in the document that the page did not already show.
              • It is drawn on the visitor's own device: the document library loads only when the button is clicked, so nothing is sent anywhere and a download spends no AI credits. If the file cannot be built, the page says so and the brief stays on screen. The shared room still gets no document of its own.

              Missing quality factors no longer borrow the demo's scores

              Engine 1.3.2, and figures moved. A firm with factors never supplied was being scored on placeholder numbers as if they were real, which inflated the result.

              • A quality factor with nothing on file now sits at a neutral 50, the midpoint of the scale, instead of inheriting the sample practice's scores. Because six factors lift a valuation together, this is a figure change, not a cosmetic one: a firm with no quality data on file now reads at a 1.0× multiplier rather than a favorable one.
              • Each factor now carries its own provenance: on file, or defaulted. The record says which, per factor, and every saved run keeps it.
              • A defaulted factor never back-solves a figure, and is never named in "What's helping", "What's holding it back", or the priced moves; a score nobody supplied is not evidence about the firm. In the book-quality table it reads "50 (default)"; "(score on file)" appears only where a score actually was.
              • The sample scores are renamed to make their one permitted use explicit: they populate a blank workbench, and are never a fallback for a stored practice.

              A figure filled from a benchmark now says so

              upsert_practice accepts an optional benchmark_derived list, so a number taken from a published median rather than from the firm travels with its citation instead of reading as owner-stated.

              • Each entry names the field (revenue, ebitda, eboc, aum, clients, cash_flow, value_per_client) and a citation that cannot be blank. The list is sent in full each time and replaces what was stored.
              • Those fields read "benchmark-derived" with their citation in the pre-flight, the workbench, and the "Where the numbers came from" section, and are excluded from the owner-stated count in the headline. Each gets its own advisory: the field was filled from a published median or assumption, not from the firm, so every method it drives is provisional.
              • The report carries the line "Filled from a published median or assumption" naming the fields and their citations. Practices with nothing recorded behave exactly as before; the column is nullable and no past figure was recomputed.

              What did not change

              • The Deal Room is app and site surface plus one guided workflow served fresh from the skills catalog. That workflow now also carries an artifact contract (skill version 1.1.0), which is read at call time rather than cached.
              • The only figure movement in this release is the neutral quality score above. The multiples, the published index, the benchmark percentiles, and the other six methods are untouched.
              • The deal room tables (memos, rooms, members, invitations, messages, and the demo cache) are new and carry their own row-level security: a memo is readable only by whoever ran it, a room only by its owner and its members, and the demo cache is not reachable from a client at all.
              • The brief PDFs are drawn in the browser from the figures the brief already shows: no server version moved, no tool changed, and a download spends no AI credits.

              One generator for the app and the chat

              Until now the Deal Room's guided workflow asked a connected assistant to write the rehearsal memo itself from the valuation report. It now calls the same server-side rehearsal the app's Deal room runs, so there is one generator and one set of rules.

              • The memo is built on the server from one saved valuation run: every figure is copied field by field from that run, along with the engine version and index version that produced it. The model writes the arguments, the answers, the terms and the checklist in words, and tags each argument with the figure it leans on.
              • An agent-run rehearsal is now held to exactly the app's rules: any line that tries to state a number is removed and the memo says how many; Jev checks each argument against the figures it cites and removes the ones they don't support; if Jev can't be reached the memo still comes back with every argument kept and the check marked unavailable.
              • The memo is saved to the same memo list as the app, as the signed-in user, and only that user can read it. The result links straight to it on the Deal room page, where it can be reopened and downloaded as the branded PDF.

              Two new tools

              Both are private sell-side prep only, neither ever reads or returns anything from a shared room.

              • `run_deal_room_rehearsal(practice_id, valuation_id?)`: runs the rehearsal on a saved run (the practice's most recent one by default). Its text is the finished memo between BEGIN MEMO and END MEMO markers, which the assistant replies with verbatim, then a one-line note with the link to the saved memo. Each run spends AI credits and only the arguments change, so the tool tells the assistant to ask before re-running when a memo already exists.
              • `list_deal_room_memos(practice_id, valuation_id?)`: your saved memos for a practice, newest first, each with the run it was built from, that run's engine version and a link to reopen it, so an earlier memo can be reopened instead of re-run.

              Deal room rehearsal skill 2.0.0

              The guided workflow was rewritten around the two tools.

              • It checks your saved memos first and offers to reopen the latest; otherwise it runs the rehearsal, presents the memo verbatim, then the What-next menu. The instruction to write the memo by hand is gone.
              • The "Deal room prep" artifact is unchanged in its sections, columns and forbidden list, and is still one static page, never a PDF or document export. Its footer now names the memo id and the saved run's date from the tool result, and it links to the saved memo in the app, which is where the PDF download lives.

              Reconnect once, and what did not change

              • The tool surface changed, from ten tools to twelve, so a client connected before this release must reconnect once to see the two new tools; `server_info` reports the live list. The in-app reconnect notice is back on for this release. Server 1.8.0.
              • The valuation engine stays at 1.3.2 and no figure moved. The shared room and the free example are untouched.

              v1.0.7

              A bigger source directory: 38 supported MCP serversSeptember 30, 2026

              Twelve new MCP servers join the supported-servers directory, taking it from 25 to 38 cards across every layer of the practice stack, firm intelligence, CRM, portfolio accounting, custody, billing, accounting and payroll. Each has its own detail page and appears in the Connect wizard, which reads the same catalog. Site content only: no tool was added or removed, no reconnect is needed, the engine stays at 1.3.1 and no computed figure moved.

              The supported-servers directory grows from 25 to 38

              Twelve new MCP servers join the directory, so every layer of the practice stack now has a connectable source. Each card carries its own detail page: hosting, authentication, the valuation fields it can supply, what you approve, and how your agent connects it.

              • Firm intelligence & registries: AdvizorPro and FINTRX join ADVmcp and AdvisorFinder as licensed cross-checks on a firm's size and headcount.
              • CRM: Wealthbox's own Connector for Claude, official hosted MCP servers from Salesforce, Microsoft Dynamics 365 Sales and HubSpot, Slant, governed access through Practifi's Sentir and NexJ, and Redtail and Advyzon reachable through Zapier's MCP endpoint.
              • Portfolio accounting & wealth platforms: Orion, Addepar, Conquest Planning, Truthifi, d1g1t, OneVest and Wealthfolio.
              • Custody & clearing: BridgeFT, Schwab Advisor Center for Claude, Charles Schwab Brokerage, and Fidelity (via Strider Labs).
              • Fee billing, accounting and payroll: BridgeFT's billing tools, QuickBooks Online, Xero, NetSuite, Sage Cloud Accounting, Zoho Books, Wave and FreshBooks, and payroll platforms such as Gusto and ADP through Zapier.
              • Orion appears in both sections it belongs to (CRM for Redtail households and portfolio accounting for assets and growth) with a card in each, both linking to the same directory page.
              • The unofficial Wealthbox and Schwab community cards now point to the official vendor connectors listed above them.

              One catalog behind both surfaces

              • The MCP servers directory and the Connect wizard read the same catalog, so every new server is offered in the connection guide the moment it is listed: no second list to keep in sync.
              • Catalog only, as always: ValuationMCP never holds provider credentials and still verifies only what lands in its own database. A server being listed says your agent can connect to it, not that ValuationMCP calls it directly.

              No reconnect, no figure change

              • The directory is site content, not the MCP tool surface: no tool was added or removed, no tool input changed, and no connected client needs to reconnect. The valuation engine stays at 1.3.1 and no computed figure moved.

              v1.0.6

              A workbench that reads as one story, artifacts that carry the house rules, and a watcher that asks the right questionSeptember 12, 2026

              Three changes, no figure moved. The valuation workbench was reorganized around the six questions an advisor actually asks, with every concept appearing once instead of up to five times. Nine guided workflows can now hand you a branded visual artifact (a pipeline board, a comparison, a diligence checklist) built from a contract that carries the same rules as the written report. And the weekly check on the published multiples stopped re-reading one page for changed numbers and started asking whether the publisher has released a newer edition, across four publishers instead of one. Server 1.7.3 → 1.7.5; the engine stays at 1.3.1 and no connected client needs to reconnect.

              The workbench reads as one story

              Reorganized around the six questions an advisor actually asks, in the order they ask them, with each concept rendered once.

              • The page used to be organized by artifact type rather than by question, so the same thing appeared several times over: the seven methods showed up in five places, the six quality factors in five, your peer position in two, and two pairs of sections shared a heading word for word. Now each idea has one home.
              • What is it worth? keeps the headline and the fair-value range, and drops the stat rows the chart below already showed.
              • Why that number? holds the plain-English explanation, the method-spread chart, and (behind disclosures, because most people never need them) the formulas, weights and confidence for each method, and the blend sliders.
              • How strong is the book? now reads as a single argument: the six factors with their scores, what each gap is worth in dollars beside it, and the priced moves to close them directly underneath. The dollar figure used to live only inside a chart.
              • How do you compare? keeps the peer chart and folds the old benchmarks panel's extra facts (your implied fee on assets, median staffing) into its footer. The InvestmentNews operating medians sit behind a disclosure, and the sourced-defaults switch moved up to the inputs where it belongs.
              • How much rides on the multiple? gathers the sensitivity view and the deal-market context under one heading, and What changed? keeps the comparison to your previous run.

              Artifacts that carry the house rules

              Nine workflows can now hand you a visual artifact, and the rules that govern the written report govern it too.

              • Each of the nine declares an artifact contract, served alongside its instructions: when to offer one, what it is called, what sections and columns it has, and what it must never show. Practice valuation, Peer position, Value driver plan, Succession readiness, Buy-sell funding gap, Acquisition pipeline, Target comparison, Diligence request list, and Deal table read.
              • An artifact is offered after the full report and after the What-next menu, never in place of either, and only when the run produced something worth keeping. A re-run with no movement does not get an offer.
              • It is layout only. Every figure on it came back from a tool in that session, unchanged: no computing, no re-rounding, no filling a blank cell. Caveats move into the footer rather than being dropped to tidy the design, and positions stay positions: no grades, no ranking, no picking a winner.
              • Each one carries its own provenance, engine version, index version, data confidence, the date it was generated, and the line saying it is an estimate for planning discussions rather than a formal appraisal or an offer. An artifact outlives the conversation and gets forwarded, so it has to argue for itself when you are not in the room.
              • Some rules are specific to the workflow. The buy-sell gap artifact names the shortfall and stops. It will not recommend a policy, a carrier or an amount of coverage. The peer position artifact is not offered at all when no cohort data came back, and every percentile on it carries its cohort definition, cohort size and filing date. The deal-table read allows no invented quotes: each seat is a lens on the computed run, not a character.
              • The look comes from the same color and type tokens the server uses to render the chart pack, so an artifact and a chart inside it are the same brand by construction rather than by two definitions that drift apart.

              Keeping the published multiples current

              The weekly check now asks whether a newer edition exists, rather than whether the numbers on one page moved.

              • These are dated publications. Re-reading the same post and comparing figures could only produce noise, and it did: a curated row meant the check flagged a change every week that was not one. The check now looks for a newly published edition instead.
              • Four publishers are watched rather than one: Sica Fletcher, whose index prices the valuation, plus Echelon, CT Acquisitions and Succession Resource Group, whose figures are shown for comparison and never priced off.
              • When a newer edition appears it is fetched, read, and staged beside the edition in use for a side-by-side comparison. Nothing switches on its own; the change is reviewed and then activated or discarded.
              • A publisher that refuses automated requests is reported as exactly that, with a date to check by hand, rather than failing quietly. If a page's layout changes so the figures cannot be read, nothing is staged and the change is flagged, better a loud gap than a quiet wrong number.
              • On its first real run the watcher found Succession Resource Group's 2026 report, one edition newer than the one on file. Its figures are staged for review; nothing shown on the site has changed until that is activated.

              No reconnect, no figure change

              • The artifact contract is an added field on responses that already existed, and the skills that carry it are database rows served fresh on every call, so every connected client sees all of it with no reconnect. No tool was added or removed and no tool input changed. Server 1.7.5; the valuation engine stays at 1.3.1 and no computed figure moved.

              v1.0.5

              Buy-side and succession workflows, acquisition screening, and a catalog you navigate by goalSeptember 11, 2026

              Seven new guided workflows (four buy-side, three succession) take the catalog from fourteen to twenty-one. The acquisition-screen skill composes advMCP's RIA Acquisition Screen with a ValuationMCP screening ceiling. The homepage now groups every skill into themed tabs (Understand, Sell, Buy, Succession, Data), and the post-valuation menu is grouped the same way. Nothing about the valuation changed: no figure moved, no engine change, no tool input changed, so a connected client needs no reconnect. Server 1.7.3; the engine stays at 1.3.1.

              Four buy-side workflows

              For a buyer or consolidator finding and sizing targets. Read from the same computed run, never a recommendation to transact.

              • Find and screen acquisition targets, give it an acquisition thesis and, with advmcp.ai connected, it drives advMCP's RIA Acquisition Screen for a ranked shortlist of named advisers from filed Form ADV data, then puts a low-confidence screening ceiling on each firm here and hands the survivors to the diligence list. advMCP's RIA Acquisition Screen now points its next step straight into this workflow.
              • The seam is stated up front: Form ADV has AUM, client counts, custody and disclosures but no revenue or profit, so every screening ceiling rests on cited-median assumptions and reports low confidence, not a valuation until the target shares its own financials. Nothing in a filing says a firm wants to sell: the shortlist is candidates to research, never a recommendation to approach, offer, or transact, and the ceiling is a walk-away, not an offer.
              • Which of these is the better buy? Two or more targets valued under identical settings and laid side by side on price, walk-away ceiling, book quality, concentration, continuity and data confidence, with the tradeoffs named and no winner picked.
              • What should I ask for before I buy? The documents and data a buyer or their counsel commonly request before closing, tailored to which of the target's figures are owner-stated versus system-sourced, and mapped to the valuation figure each one would confirm. A request list, never legal or M&A advice.
              • What does this add to my book? A target read against the buyer's own practice as a combined position: blended AUM tier, how concentration and client mix shift, where the two books overlap. Arithmetic on two computed records, never a projected synergy or a combined price.

              Three succession workflows

              For an owner planning the handoff: the continuity picture already inside the valuation, with structural questions routed to the owner's attorney, CPA, and insurance advisor.

              • What to put on paper before you hand it over: the document-level moves that raise transferability (a named successor, a written plan, a funded buy-sell), each with the engine's own dollar effect, because these are paperwork rather than business change.
              • Does your buy-sell still cover the number? An existing buy-sell's owner-stated funded amount compared to the practice's current value, the gap named in dollars and routed to the attorney and a licensed insurance advisor. It surfaces the gap; it never recommends coverage or designs the agreement.
              • What does this number mean for an internal buyout? The fair estimate and range read for a G2 or partner buy-in: what tends to be financeable versus seller-carried, and the questions that set the terms. It describes what the number implies and never designs the deal.

              A catalog you navigate by goal

              Twenty-one skills is too many for one flat list, so the catalog now groups by what you're trying to do.

              • Every skill carries theme tags (understand, sell, buy, succession, data) returned by both list_skills and get_skill, so an agent can filter the catalog the same way the site does.
              • The homepage catalog is grouped into themed tabs with a search box above them; search spans every theme so it never hides a match. The active tab and the search term live in the URL, so a themed view is a shareable link.
              • The post-valuation "What next?" menu is grouped by the same themes, recommended workflows first, with the new buy-side and succession workflows suggested only when the run gives a concrete reason: a buy-side lane, an unfunded buy-sell, a near-term retirement.
              • Switching a tab or typing in search now holds your place on the page instead of jumping back to the top.

              No reconnect, no figure change

              • The seven new skills are database rows served fresh on every call, and the theme tag is an added field on responses that already existed, so every connected client sees all of it with no reconnect. Server 1.7.3; the valuation engine stays at 1.3.1 and no computed figure moved.

              v1.0.4

              Engine 1.3.0: the % of assets method prices off a published rateSeptember 7, 2026

              Figures moved. The AUM Percentage method previously applied a 1.5% rate written into the engine; it now uses the published Sica Fletcher AUM percentage for the practice's size tier (midpoint), so every valuation's % of assets method changes, down for practices under roughly $225M of assets, up for larger ones. Runs computed under engine 1.2.0 are not comparable to 1.3.0 runs.

              The % of assets method now uses a published rate, figures moved

              Engine 1.3.0. Until now the AUM Percentage method applied a flat 1.5% of assets that was written into the engine, while the method carried the second-highest confidence in the blend. Sica Fletcher publishes AUM percentages by firm size on the same page the index already cites, so the method now reads them instead. Every valuation's % of assets figure changes, and with it the blended fair value.

              • The published "Assets Under Management (AUM) Percentages for RIAs" table joins the index as a third axis (keyed on assets, alongside the revenue and EBITDA axes) at index version 2: the same 2024 edition, same publisher, additional data. Because the engine does not know a practice's RIA type, the five published types become the width of the range: small 0.9–1.3% (midpoint 1.10%), mid 1.4–1.7% (1.55%), large 1.9–2.3% (2.10%).
              • Each practice resolves the midpoint for its size tier (small below $225M of assets, mid up to $1.6B, large above) following the same tier-cut convention already used for revenue and EBITDA. Practices under roughly $225M move down, larger practices move up.
              • Confidence is now earned rather than assumed: 88 when the rate came from the published index, 60 when the practice falls below the smallest published anchor or no tier applies and the engine constant is used: the same treatment EBOC gets when it runs on a placeholder. In that case the run says so in its own words and lists the rate among the figures no publisher supplies.
              • The 1.5% constant survives only as the out-of-tier fallback, renamed so it cannot be mistaken for a sourced figure, and the method note now states which basis it actually used.
              • The sensitivity strip sweeps the published AUM range along with revenue and EBITDA, so it reflects the full published spread instead of understating it. Owner-supplied figures stay fixed as before.
              • Runs computed under engine 1.2.0 are not comparable to 1.3.0 runs. Valuation Delta already treats a different engine version as non-comparable and offers a matching rerun. No saved run was altered.
              • Reference figures for the sample practice: fair value $5.63M to $5.67M, asking price $6.20M to $6.23M.

              v1.0.3

              Advisor-first report: index multiples, verbatim ladder, workbench links, peer cohort, and published cross-checksSeptember 6–7, 2026

              The September 6–7 consolidation. The engine prices off the published Sica Fletcher index for the firm's size tier, versioned, watched weekly for a new edition, and stamped on every run. Reports are server-rendered as a plain-English ladder with verbatim markers, source attribution, and a workbench deep link. get_benchmarks folds in an ADVmcp peer cohort and now also returns other publishers' multiples for comparison only, three new guided skills join the catalog, and the homepage source cards build themselves from the live data. Only figure change: the reference fair value moved from $5.36M to $5.63M under index multiples.

              Multiples: one sourced index prices the number, every other publisher is shown beside it

              Everything about multiples in this release, in one place. The engine prices exclusively off the versioned Sica Fletcher SCF Index for the firm's size tier; other publishers' figures are carried for comparison and can never reach a calculation. Owners are never asked for multiples. Engine version stays 1.2.0.

              • Revenue and EBITDA multiples default to the index midpoint for the practice's size tier; Comparable Transactions prices off the tier low times the quality multiplier; EBOC keeps a flagged placeholder at reduced weight. The result carries resolvedMultiples and multipleSensitivity, and saved runs stamp the engine version, so Valuation Delta treats an engine change as non-comparable.
              • The published multiples are a sourced, versioned dataset rather than a constant. The Sica Fletcher SCF Index ranges (2024 data) and the Echelon Partners deal-activity figures live in versioned tables, and every surface (get_benchmarks, the engine, and the workbench) reads the active version through one loader; the built-in copy is only an offline fallback, and the response says which was used.
              • Every Monday 09:00 ET the publisher pages are re-read. An unchanged page only refreshes the verified date; a changed or unreadable page is staged for review in the admin "Index updates" panel, where the owner can diff active vs staged and Activate or Discard. Nothing switches automatically and every check is logged.
              • Saved runs record the index version they priced off. Valuation Delta treats a different index version like an engine change, not like-for-like, and offers a matching rerun. Existing runs were stamped index v1.
              • The source is shown wherever the number appears: the workbench headline links to the index with its as-of and verified date, the chat headline card names the published index and its verified date rather than implying deal comps, and the sensitivity strip cites the source.
              • Owners are not asked for multiples anywhere. Every skill, the compute_valuation report, warnings, engine notes, the get_benchmarks caveat, the server instructions, and the workbench copy were swept of lines that sent an owner to find multiples. The engine supplies them and says so; multiples an owner volunteers are accepted and labeled; any adjustment is a workbench choice with a "Reset to index" affordance, never a task. EBOC keeps its placeholder flag, framed as the system's gap, not the owner's. A test fails the build if the retired phrasings reappear.
              • get_benchmarks also returns every other published multiples source that matches a practice, under `crossCheckMultiples`: CT Acquisitions' 2026 RIA M&A ranges tiered by AUM band, and Succession Resource Group's 2025 Advisor M&A Report averages across 176 deals completed in calendar 2024 representing over $13.3B in assets. Display only: shown for comparison in *What's the market doing?*, never an input to a calculation, and the table refuses to be a pricing input at the database level so a future change that tried to wire it into the engine fails there rather than quietly moving everyone's valuation.
              • Those cross-check figures are kept in their own versioned table because they are tiered on assets under management rather than on firm size by revenue and profit, mixing the two axes would corrupt both. A source that publishes averages across all firms rather than ranges by size is flagged as not size-matched and labeled "all firms" wherever it appears. Reviewable in the admin panel alongside the index and market-activity datasets; the weekly automated check still covers the index and market activity only.
              • *What's the market doing?* moves to 1.2.0 with a "second opinion on the multiples" section: every matching source with its publisher, that publisher's side of the market and its sample; size-matched bands kept visually separate from all-firm averages; what a range would do to this practice's own figure as plain arithmetic; where the pricing index sits among the published sources; and a closing reminder that the number did not change. Six new guardrails: never blend or split the difference between sources, name the axis and size-matching differences first when they disagree, name each publisher's side of the market, never present a cross-check as deal comparables or as equal in standing to the index, never call the index too low or the estimate understated, and never present an all-firm average as what firms this size fetch.
              • The homepage source cards now build themselves from the live tables and are grouped by role (what sets the multiples, what positions the practice, and what is carried for comparison only) so a newly added source appears on its own.
              • Two standing context sections appear on every first screen: How you compare to firms like yours, and The market you'd be selling into. Workbench and compute_valuation share one compute path and one default (confidence weighting), and a parity test pins both surfaces to the same figures. No valuation figure changes.

              Peer cohort folded in

              Cohort context arrives with the benchmarks, so an assistant does not need a second server connected to get it.

              • get_benchmarks now folds in an SEC Form ADV peer cohort from ADVmcp under advCohort (AUM band, optional state), so cohort context is available without connecting advmcp.ai.
              • The cohort is best-effort: if ADVmcp is unreachable or returns a cohort that does not match the requested band, advCohort reports available: false with a reason and every other benchmark field returns unchanged.
              • Clients connected before this release are working from a cached tool list and will not see the new `state` input until they reconnect the server.
              • The cohort lookup gets 8 seconds rather than 3.5 before it is given up on, and reports how long it took and why it failed, so a slow-but-healthy lookup lands instead of silently reading as unavailable.
              • No engine or figure changes; engine version stays 1.2.0.

              Three new guided skills

              Served fresh from the skills catalog, so they reached every connected client with no reconnect. Each closes with the same plain-English ladder as the existing workflows.

              • Reconcile the Numbers: walks a figure that looks wrong back to the field, the source that supplied it, and the method it drives, so a disputed number is settled against the record rather than re-argued.
              • Transition Risk Read: reads what happens to the practice if the owner steps away: successor presence, plan and funding, owner-held relationships, and advisor tenure, priced through the longevity factor.
              • Market Context: puts the run in the market it would sell into, using the cited index range for the firm's size tier and Echelon Partners' deal-activity context, never as a deal comparable.
              • The What-next menu after a valuation now stars a skill conditionally on the run's own signals (plausibility findings, engine-default placeholders, size-tier mismatch, and longevity gaps) so the suggested next step follows the data rather than a fixed order.

              Open this run in the workbench

              Every valuation result carries a deep link to the same practice and saved run in the web app, so the chat and the workbench are two views of one number.

              • compute_valuation returns structuredContent.workbenchUrl (practice and, for saved runs, the run id); list_practices carries the same URL on every history entry. The workbench accepts the link, signs you in if needed, preselects the practice, and loads that run's inputs and weights. Unknown ids fall back to the normal workbench with a notice. Server 1.2.2.
              • In the report the link sits directly under the firm name, with the address behind the word workbench, so it is the second line an advisor reads. Server 1.2.4.

              Verbatim by construction

              Consuming assistants were condensing the report. It is now the tool's only text output, bounded by markers, so there is nothing else to summarize.

              • compute_valuation's text is exactly: an instruction line, the report between === BEGIN REPORT === and === END REPORT ===, then the What-next menu. The JSON payload left the text and lives only in structuredContent, which also carries the bare report and the instruction. validate_practice follows the same convention. Server 1.2.3.
              • The instruction to reply with the marked text verbatim is now the first line of the server's own instructions (loaded by every client at connection), and the presentation-contract and practice-valuation-run skills say the same. A test runs the real handler and asserts one text item, no JSON, one marker pair, and report-in-text equal to structuredContent.report.
              • Integrations that parsed the JSON out of the text need to read structuredContent instead; that is the only breaking change in this group.

              Sources and completeness in the readout

              The confidence phrase gets its number back, and the report names every system that supplied a figure.

              • The headline reads, for example, Built on a complete set of inputs (14 of 14, 100%). A new section, Where the numbers came from, lists each connected source with the plain-language fields it supplied and its last sync date, any owner-stated fields no platform covered, engine-derived fields, provenance conflicts (you said X, the platform reports Y, the report uses which), and the benchmark sources with their as-of dates and cohort size. It closes with the completeness count, the engine version, and the index version. validate_practice renders the identical section, so pre-flight and readout agree. Server 1.2.5.

              A layout that reads

              Same words, real structure: headings, rules, and one idea per paragraph, rendered correctly by chat clients.

              • Firm name as a heading, then the workbench link, then an Asking price heading with the fair estimate and range on one line. Sections are headings, the two data blocks are sub-headings, a rule separates each major part, What this means is four short paragraphs, and every table, list, and rule is surrounded by blank lines. The pre-flight report uses the same rules; a layout test enforces them, and the skills now tell the assistant to preserve headings, blank lines, rules, and tables exactly. Server 1.2.6.

              Plain-English ladder with progressive disclosure

              Owners and advisors are the primary readers.

              • Every workflow skill renders the same ladder: headline card, What this means, What's helping and holding it back with dollar figures, How to move the number, Before you rely on this, and a three-item What next. Each layer ends with Show me the math, which expands the full tables for buyers, bankers, and consultants. Skills use plain titles and a confidence phrase. Server 1.0.11.
              • The skills table gained a committed migration and seed file as its source of truth.

              The server writes the report

              After a rerun collapsed to a one-line result in production, the entire ladder is rendered server-side so no assistant has to reconstruct it.

              • compute_valuation returns the complete report in its text and again as structuredContent.report, written in the plain register from the engine's own explainability fields. A rerun is a full report; Since your last run is one line inside it, never a substitute. Raw quality metrics are now persisted on the practice so the fact behind each score reads as a real fact. Server 1.1.2.

              v1.0.2

              Drift detection, deterministic pre-flight, currency, presentation contract, structured output contracts, Valuation Council, and cited transaction-multiple rangesSeptember 2–5, 2026

              The September build wave. September 2 added the server_info drift detector, fair-use rate limits, engine-version stamping, sanitized errors, and the benchmark-against-cohort skill. September 4 added validate_practice (a deterministic pre-flight the assistant runs before computing), owner-supplied currency handling, a house presentation contract every skill renders by, two pre-flight fixes, and a machine-readable output contract on every skill. September 5 added the Valuation Council skill, cited transaction-multiple ranges (Sica Fletcher, Echelon Partners), and a guided next-step menu after every valuation. No valuation figure changes.

              Self-healing update surface

              MCP clients cache the tool list when the connection is created, so an older connection can silently miss newer tools. The server now makes that detectable.

              • `server_info`: reports the server version, the valuation engine version, the live registered tool list and count, the skills catalog size and last update, and a cache note explaining how to refresh a stale connection. Read-only, no authentication needed.
              • The tool list in `server_info` is read from the tools the server actually registered at runtime, so it can never drift from a hand-maintained list.
              • Server instructions now tell agents to compare their available tools against `server_info` and, on a mismatch, tell the user their client cached an old tool list.
              • New skills need no reconnect: `list_skills` and `get_skill` always serve the latest versions from the database.
              • `server_info` is callable without authentication, so status pages and monitoring can read the server's live state.
              • `benchmark-against-cohort`: places a practice against filed SEC Form ADV peer percentiles while valuing it, with cohort data from advmcp.ai used to sanity-check the figures a valuation was built from and to place the result in context, never to move the number.
              • This skill needed no reconnect: skills are served fresh on every call, so a new guided workflow reaches every connected client the moment it is published; only the `server_info` tool required the one-time reconnect.
              • Tool count grows from eight to nine; all previous tools are unchanged.

              Reliability & trust hardening

              The public MCP endpoint and saved valuation records are now more robust, reproducible, and safely framed.

              • A fair-use rate limit now protects the public MCP endpoint (per-caller minute, hour and day windows, with a retry-after on 429s). It fails open if the counter store is unreachable, so the service degrades gracefully rather than going dark.
              • Every saved valuation now records the deterministic engine version that produced it (engine v1.0.0), so any past run can be reproduced and audited even after the engine's math evolves.
              • Valuation summaries now close with an explicit framing line: educational estimate from the deterministic valuation engine, not an appraisal, fairness opinion, or investment, legal, or tax advice.
              • Tool errors now return a short generic message instead of raw database detail, with the real error logged server-side.

              Reconnect instructions

              The in-app reconnect notice is now a two-line nudge that links here; the full reset steps live below.

              • Cursor: Settings → MCP, toggle ValuationMCP off and back on (or hit refresh on the server entry).
              • Claude: remove the ValuationMCP connector and add it again with the same URL.
              • Other clients: disconnect and reconnect, or restart the client so it re-runs tool discovery.
              • The MCP endpoint is unchanged, and no new account or key is needed.

              New tool: validate_practice

              A deterministic pre-flight, run before compute_valuation. Nothing here is model-computed.

              • Returns each present field's provenance (system-sourced vs owner-stated), each missing field and its exact point cost toward data confidence, which engine-default multiples are silently in play, any place an owner-stated figure diverges from what a platform actually pulled, currency status against the USD benchmarks, and revenue- and AUM-per-household plausibility against the tier, split into blocking and advisory findings.
              • compute_valuation now always attaches the same pre-flight to its result and records an explicit gaps_acknowledged flag, so gaps cannot be skipped, but it never blocks a knowing partial run.
              • record_data_source now stores the actual value each platform returned, so a later run can flag when an owner-stated figure diverged from what the platform pulled.
              • Tool count grows from nine to ten. This is a tool-surface change, so a client connected earlier must reconnect once to see validate_practice; server_info reports the live count.
              • The Practice Valuation Run skill gained a validate step between the benchmark sanity-check and compute.

              A house presentation standard, in prose and in structure

              Every guided workflow now renders the same way, enforced two ways: a prose contract the assistant reads, and a machine-readable contract a client can render. Layout only; it never changes a computed figure.

              • The presentation-contract skill sets three rules: lead with the answer (value, range, confidence together), never buried in a paragraph; put any set of three or more parallel figures (the method spread, quality factors, benchmark positions, data gaps) in a table rather than a comma-run; and collect every caveat into one closing block. Engine-default and cited-median figures are flagged inline in their own cell.
              • Alongside the prose rules, every skill now carries a machine-readable output_contract (a format plus the ordered sections its answer should follow (or a style/principle for the reference skills)) returned by both list_skills and get_skill, so a client can render the structure directly.
              • Wired into all eight guided workflows and served fresh from the database, so it reached every connected client with no reconnect.

              Pre-flight fixes

              Two issues in validate_practice, surfaced by real use.

              • Quality-factor provenance: the six quality factors now credit the platforms their raw metrics came from. They previously read "owner-stated" even when a platform supplied them, because provenance was keyed on the derived score name rather than the raw metric name recorded against it.
              • Data confidence: the server-derived value_per_client no longer counts as a missing field, so a fully-sourced practice reads 100% coverage and clears the "provisional" flag on its own, and a phantom "close data gaps" to-do for a field the server fills in itself is gone. No valuation figure changes: only the reported confidence signal.

              New skill: Valuation Council

              A completed valuation argued from every seat at the deal table, with a neutral moderator: the seven-advisors council pattern applied to a valuation.

              • Eight seats, each a lens on the same computed run rather than a character with invented figures: Seller's Broker, Buyer, Lender, Practice Manager, Succession Planning, Business Partner, a CPA after-tax lens, and The Appraiser's Bar.
              • A neutral moderator frames the question the table is convened on, keeps every seat on the computed numbers, names where they agree and diverge, prices the exercise by data confidence, and routes to a credentialed appraiser, CPA, or attorney on a live deal, but never decides or recommends.
              • The Appraiser's Bar describes what a credentialed appraiser (CVA, ABV, CBA) would require before signing and when to hire one; it never roleplays the credential or issues an opinion in one's name.
              • Served fresh from the database, so it reached every connected client with no reconnect.

              Cited transaction-multiple ranges

              Transaction multiples were the engine's weakest input, silent defaults with no published source behind them. get_benchmarks now carries cited aggregate ranges.

              • get_benchmarks returns a transactionMultiples block: revenue and EBITDA ranges by firm-size tier from Sica Fletcher's SCF Index (2024 data), plus market-activity context from Echelon Partners' 2025 RIA M&A Deal Report: each with source, as-of, and verified date.
              • Sica Fletcher publishes tiered multiple ranges (the multiples source); Echelon Partners publishes deal activity, not standardized multiples, so it is cited as market context, never as a median multiple.
              • The validate_practice pre-flight now appends the cited size-tier range to the revenue- and EBITDA-multiple "engine default in play" advisories, chosen from the practice's own figures; EBOC and the fixed comparable have no published range and say so.
              • Ranges are labeled size-sensitive index figures, not named deal comparables, and are never applied to the valuation: the engine runs on its defaults until you set a multiple from deal data you trust, adopted via upsert_practice and logged as a source.

              Guided next step after every valuation

              A completed valuation now ends with a menu of what to do next, so the workflow doesn't dead-end at the number.

              • `compute_valuation` returns a `nextSteps` menu: the applicable guided skills, each with a one-line "when to use" and a ready-to-send prompt, and renders it as a "What next?" list at the end of the result.
              • The menu is context-aware: Data Confidence Triage is starred when confidence is low, Valuation Delta activates only once a second run exists, and Buy-Side Ceiling is flagged as a separate lane.
              • Served live from the skills catalog in the home-page order, so new skills and any reordering appear in the menu with no code change and no reconnect.
              • The Practice Valuation Run skill now closes by surfacing that menu and running whichever option the user picks, layout and routing only; it never changes a figure or auto-runs an unchosen skill.
              • The same menu now appears in the web app: a "What next?" grid in the workbench after a run, and a per-run "What next?" panel on the recorded-valuations page, each with one-click Copy-prompt buttons to paste into your agent.

              v1.0.1

              Guided skills and reconnectAugust 26, 2026

              Adds a public skills catalog with two new MCP tools, surfaces it on the homepage, and prompts clients connected at v1.0 to refresh their cached tool list.

              Guided skills catalog

              A public, read-only catalog of step-by-step workflows an agent can discover and follow. No account or key required.

              • `list_skills`: browse the catalog of guided skills with titles, blurbs, kinds, versions, and sort order.
              • `get_skill`: load a single skill by slug so an agent can follow a written procedure for a valuation task.
              • Server instructions now tell agents to check the skills catalog first, so a connected client follows the published procedure instead of improvising one.
              • Tool count grows from six to eight; all six v1.0 tools are unchanged.
              • 32 guided skills, read live from the catalog:
              • Valuation Literacy: How the seven-method blend actually works: what drives each method, what the quality multiplier touches, what data confidence measures, and the vocabulary that keeps a valuation honest.
              • Presentation Contract: The house rules for rendering any valuation output: lead with the number, put parallel figures in tables, and collect every caveat at the end. Layout only; it never changes a figure.
              • Explain my valuation: A valuation consultant in your workbench who explains what your asking price, range, coverage and methods mean, and why the engine calculated them that way, in words anyone can follow, then points you to the next best step.
              • ValuationMCP Agent: The owner's category expert driving assigned moves through tasks, sourced analysis and draft deliverables, with owner-approved decisions.
              • Value my practice: The full sell-side loop: requirements, gathering from the owner's own platforms, provenance, benchmark sanity-checks, and a confidence-weighted blend presented with its caveats in the right order.
              • What's missing, and is it worth getting?: Turns a confidence percentage into a to-do list: which missing facts cost the most points, which platform holds each, and what the quality factors are silently assuming in the meantime.
              • Do these numbers hold up?: The figures behind a valuation, checked against filed data and published medians: where they disagree, what could explain each gap, and where to go and settle it. Questions for the owner, never corrections to the model.
              • What's holding my score back?: Which of the six quality factors is moving the multiplier, by how much, and what the arithmetic says a different score would do: sensitivity as math, never as advice to change the business.
              • What happens if you step away?: The continuity picture already inside your valuation, read as one story: who owns the relationships, what is on paper, and how much runway there is. Plus an honest account of how little of it this model actually prices.
              • What to put on paper before you hand it over: The document-level moves that raise transferability (a named successor, a written plan, a funded buy-sell), each with the engine's own dollar effect, because these are paperwork rather than business change and the cheapest lever an owner has.
              • What does this number mean for an internal buyout?: The fair estimate and its range read for a G2 or partner buy-in: what portion tends to be financeable versus seller-carried, and the questions that actually set the terms. It describes what the number implies, never designs the deal.
              • Does your buy-sell still cover the number?: An existing buy-sell agreement's funded amount, compared to the practice's current value: the gap named in dollars, with the structural questions routed to your attorney and insurance advisor. It surfaces the gap; it does not solve it.
              • Find and screen acquisition targets: Turn an acquisition thesis into a ranked shortlist and a screening ceiling for each firm. advMCP finds the candidates from filed Form ADV data, and ValuationMCP puts a low-confidence walk-away number on each. A research list, never an approach list.
              • How much should I pay?: A buyer-side ceiling estimate: given the practice's economics and the market dataset, what price would leave you with a defensible return?
              • Which of these is the better buy?: Two or more practices you might acquire, read side by side (fair estimate, walk-away ceiling, book quality, concentration, continuity and data confidence), with the tradeoffs named and no winner picked.
              • What should I ask for before I buy?: The documents and data a buyer or their counsel commonly request before closing on a practice (financials, filings, client agreements, custody, comp, concentration and continuity) as a request list, never a to-do the model priced.
              • What does this add to my book?: A target read against the buyer's own practice (the combined AUM tier, how concentration and client mix shift, the blended revenue per client, and where the two books overlap) as the combined position from figures already computed, never a projected synergy.
              • Request data from a target: Send a target firm a private, voluntary data request from your assistant, with a dry-run preview and an email left as a draft.
              • Read a target's retirement plan book: See which retirement plans name a buy target as adviser on Form 5500 Schedule C, and turn that into diligence questions.
              • What's the market doing?: The published multiples your valuation priced off: which size tier you fall in and why, what the whole range does to your number, which two of the seven methods run on figures nobody publishes, and how current the data is.
              • Answer a data request: Answer a received information request from the firm's own records, with a preview and explicit owner approval.
              • What does this number mean?: Reads a completed valuation the way a seasoned practitioner would: why the methods disagree, which number each side of a deal anchors on, what to probe next, and when the situation has outgrown a modeled estimate.
              • Deal room rehearsal: Private seller or buyer preparation: rehearse the other seat’s arguments, your answers, terms and verification checklist from one saved run.
              • Plan my moves: Plan my moves (formerly "How do I move the number?"): the engine's priced driver block turned into moves the owner can plan, track and prove. What each is worth, the arithmetic from your inputs, the first step and how you'll know.
              • Prove a move: For a move you marked done: record the new figure with where it came from, re-run, and see whether the server verified it on file.
              • Move check-in: Where your plan stands: overdue and due-soon moves, stalled steps, moves waiting for proof, and whether a re-run is due.
              • Billing: fees, recurring mix and concentration: For the revenue-per-household move: pull your household billing from a connected server (or an export) for yield bands, households below a fee floor and discounts against your stated schedule, then price a floor as a revenue scenario.
              • Client book: For the retention, concentration and demographics moves: pull your households from a connected CRM (or an export) for retention over the last year, an at-risk review list, top-ten dependency on single advisors, the age and life-stage mix and next-generation coverage.
              • What changed since last time?: Compare two saved runs and attribute the movement (business figures, data completeness, assumptions, or settings) so a change in modeled value is never mistaken for a change in the market.
              • Close the margin gap: Work the operating-margin move: pull the P&L from the owner's own accounting server, show where margin is lost by expense category against the size-tier median, draft plan steps, and save only the totals the owner confirms.
              • How do I compare to firms like mine?: The first screen already shows your position on three metrics; this skill goes deeper, with state and size cohorts, more metrics and the distribution chart.
              • Argue it from every side: A completed valuation argued from every seat at the deal table (seller's broker, buyer, lender, practice manager, succession planning, business partner, the CPA's after-tax lens, and the bar a credentialed appraiser would set), with a neutral moderator who surfaces where the seats agree, where they diverge, and where the situation has outgrown a modeled estimate. Each seat is a lens on the same computed number, never a character with invented figures.

              Homepage skills surface

              The catalog is visible on the marketing site, not just over MCP.

              • Live skills section on the homepage, loaded in the route loader from the database so the catalog stays current without a deploy.
              • Dedicated icon for each skill, matched to its slug.
              • Tools grid showing all eight MCP tools with what each one does.
              • A fetch failure degrades to an empty catalog rather than breaking the page.

              Reconnect notice

              MCP clients cache the tool list at connection time, so a client connected at v1.0 never sees the two new tools.

              • Dismissible slide-in notice explaining the caching behavior and the six-to-eight tool change.
              • Reset instructions for Cursor, Claude, and other MCP clients.
              • The MCP endpoint is unchanged and no new account or key is needed.
              • Shows once per browser and links to these release notes.

              Data access refinements

              • `list_practices` returns full valuation history (label, notes, weights, inputs, quality, methods, quality multiplier, and prior-run link) for the 24 most recent runs when a practice id is supplied.
              • Legal entity updated to HnyComb.ai LLC across the Terms of Service and Privacy Policy, with contact routed to hnycomb.ai.

              v1.0

              First public releaseAugust 12, 2026

              The initial ValuationMCP release: the MCP server, the seven-method valuation blend, SEC Form ADV benchmarks, explainability and optimization, accounts, admin oversight, and the privacy posture.

              MCP server

              The primary interface to ValuationMCP. Agents connect over Streamable HTTP at /mcp and authenticate with OAuth 2.1, so every tool call runs as the signed-in user.

              • OAuth 2.1 authorization with dynamic client registration, an in-app consent screen, and per-user scoping on every read and write.
              • `list_practices`: enumerate the practices on the account with current data coverage.
              • `upsert_practice`: create a practice or merge newly gathered facts into one; omitted fields are left untouched.
              • `list_data_requirements`: the full checklist of facts a valuation needs, which method each fact drives, and which kind of platform normally holds it.
              • `record_data_source`: log provenance: which platform supplied which field, and when.
              • `get_benchmarks`: sourced SEC Form ADV percentiles and published operating medians for sanity-checking gathered figures.
              • `compute_valuation`: run the blend and save the result, returning every method's arithmetic, the quality multiplier, the weighted blend, the headline, explainability narrative, optimization recommendations, and prior-run comparison.
              • Raw-metric intake: recurring revenue %, organic growth %, retention %, average client age, top-10 revenue share %, and advisor tenure are converted to the six 0–100 quality scores server-side with fixed, published anchors, so identical inputs always score identically across runs and clients.
              • Composite quality factors: `quality_longevity` is a succession-readiness score weighted across successor presence, written plan, funded buy-sell, owner relationship share, advisor tenure, and years to retirement; `quality_demographics` blends average client age with the asset-weighted decumulator share.
              • Server-side derivation of `value_per_client`: capitalizes revenue per household at the firm's revenue multiple, falling back to the tier median when the firm's own figures are missing, so agents no longer invent a capitalization factor.
              • Server instructions that steer agents away from inventing figures: a missing field lowers reported data confidence instead of producing a confident-looking guess.

              Valuation engine

              Seven methods, each reported with the arithmetic behind it, blended by a normalized weighted mean.

              • Revenue Multiple, EBITDA Multiple, EBOC Multiple, Discounted Cash Flow, Client Value, AUM Percentage (1.5% base × quality), and Comparable Transactions (2.2× revenue × quality).
              • Six-factor quality multiplier (recurring revenue mix, organic growth, client retention, client demographics, concentration, and succession longevity) clamped to 0.5×–1.5×, with each factor shown at its point of effect.
              • User-adjustable method weights plus a "weight by data confidence" preset that loads each method's own confidence score into the weight vector (same normalized-mean math, different weights).
              • Per-method data-confidence scoring and a weighted coverage score so a thin dataset produces a visibly low-confidence valuation.
              • Sell-side headline: a defensible suggested asking price set above the fair-value blend, with the low/high range across contributing methods.
              • Client Value method can use a server-derived value per client when the user does not supply one.
              • Manual Workbench for exploring the same math by hand, available to signed-in accounts.
              • Deterministic insights engine explains what drove the valuation: revenue scale, quality multiplier, method dispersion, and data caveats.
              • Practice-management optimization recommendations are priced by re-running the blend with the improved metric, so each suggestion shows an estimated uplift value.
              • Recommendations cover recurring-revenue conversion, succession planning, concentration reduction, client-demographics improvement, and retention programs.
              • Automatic comparison to the most recent prior valuation for the same practice: changes in asking price, fair value, quality multiplier, and individual method values.
              • Insights are saved with every recorded valuation and surfaced in the workbench, the history page, and the MCP `compute_valuation` response.

              Practice workflow

              The workspace centers on choosing a firm first, then deciding what to do with it.

              • 'Your practices' section so signed-in users can select an existing firm or create a new one on login.
              • Inline Rename and Delete actions for practices in both the client workspace and the admin dashboard.
              • Cascading delete: removing a practice also removes its recorded valuations and data-source provenance logs.
              • Post-selection panel shows the selected firm and offers two clear paths: view the data already on file, or gather new data.
              • Inline data viewer displays 'On file' fields formatted as USD or percent, plus 'Still missing' fields with the platform that normally supplies each one.
              • Quick links from the selected firm to its valuation history and the manual workbench.
              • Practice dropdown at the top of the workbench loads every firm on the account and auto-populates the calculator with its saved inputs and quality scores.

              Benchmarks

              Only sourced figures, no invented industry data.

              • SEC IAPD Form ADV panel derived from the official SEC compilation feed covering 23,512 registered investment adviser firms, with firm-size and client-mix percentiles.
              • Published industry operating medians (revenue per client, operating profit margin) used for tier-aware starting points.
              • Transaction multiples are explicitly left as user-editable inputs, because neither benchmark source publishes deal multiples.
              • Benchmark and industry panels surface the source and attribution next to every figure.

              Accounts and workspace

              Registration gates the MCP server and the valuation surfaces.

              • Email and Google sign-in, with a workspace portal that shows connection status and recorded valuations.
              • Connection guide mapping wealth-tech platforms (CRM, portfolio accounting, custodian, billing, accounting, payroll) to the exact fields they supply.
              • Supported MCP servers directory with a dedicated page per server covering hosting, authentication, and which valuation fields it can feed.
              • Recorded-valuation history with labels, notes, provenance summary, explainability, optimization, prior-run comparison, and delete.

              Admin

              Output-only oversight across accounts.

              • Valuation ledger showing each account that ran a valuation, the provenance of its data, and the calculated results, never the input data.
              • Cross-firm averages across every valuation calculated in the app.
              • Management of recorded valuations, and a contact-form inbox with reply and delete.
              • Client/Admin mode switch inside the logged-in workspace.

              Brand, site and privacy

              • ValuationMCP identity on valuationmcp.ai: light and dark logo treatments, favicon and touch icon, blue-violet brand palette and gradients drawn from the mark.
              • Dark mode by default with a light/dark toggle that persists.
              • Landing page positioned as an MCP-native valuation app built for wealth management, with a standardized footer across every page.
              • Contact form that routes inquiries without exposing an email address.
              • Privacy policy documenting what we store: the practice figures your agent supplies through `upsert_practice`, their provenance, and the calculated valuation outputs. We never copy the underlying records, exports or files from connected platforms, and we never hold your platform credentials. Rights are limited to aggregated, de-identified benchmarks. (Corrected 2026-08-18: this entry previously described raw inputs as never stored; practice figures supplied through `upsert_practice` are stored on the practice record; see the Privacy Policy.)

              Known limits

              • Buy-side and succession workflows are now in the catalog, but they read the same computed run and public or target-provided data. They are estimates and read-only, never a recommendation to approach, offer, or transact.
              • Deal room memos and shared-room agendas are a rehearsal written from a computed valuation: they state no figure the engine did not return, and they are an educational exercise, not an appraisal, fairness opinion, or negotiating advice. The free example prices a firm from public filings alone, so its revenue and profit are median-filled estimates and every method they drive is provisional.
              • A rehearsal run from a connected assistant waits for the whole memo before it replies, which can take a little while; the tool sends no progress updates yet, so a client with a short timeout may give up first; the memo is still saved and can be reopened with list_deal_room_memos or in the app's Deal room.
              • Transaction multiples come from a published size-tier index (Sica Fletcher SCF Index), versioned, with the current edition watched weekly for an update; they are aggregate ranges for firms of that size, not deals for this practice, and can be adjusted in the workbench. EBOC has no published index and uses a flagged placeholder.
              • Data collection is the agent's job and can run long across multiple source platforms; each MCP call itself is a fast, synchronous request. Scoring, the seven-method blend, and explainability/optimization all execute server-side inside ValuationMCP.

              Browse the 62 supported MCP servers across 9 source categories or read the privacy policy.