ValuationMCP
Start free

Privacy Policy

Effective date: October 19, 2026

Last updated: November 13, 2026

1. What ValuationMCP is

ValuationMCP is an MCP-native, A2A-compatible valuation application for wealth-management practices. Your AI agent connects to our MCP server (or our agent-to-agent endpoint), reaches into the platforms that already hold your practice data (such as your CRM, portfolio accounting system, custodian, billing tool, or general ledger) and sends the figures it finds to our application, which scores them, runs the valuation methods, and stores the result.

The application has two lanes. The sell lane values your own practice for planning, and can open shared Diligence Rooms for invited counterparties. The buy lane tracks acquisition targets you are evaluating privately, and lets you request diligence figures from those firms through data requests.

2. What we store

When your agent calls the upsert_practice tool, or when you enter the same figures yourself in the web workbench, the practice inputs are saved to a practice record in our database. That includes the firm name and any free-text notes you or your agent provide, along with the financial and operating inputs sent:

  • Revenue (trailing 12-month), EBITDA, and EBOC (earnings before owner compensation)
  • AUM and client household count
  • Cash flow used as the DCF base year, growth rate, discount rate, and terminal growth rate
  • Revenue multiple, EBITDA multiple, EBOC multiple, and value per client
  • The firm name and any free-text notes about the firm or the data gathered

These practice records persist until you delete them. They are scoped to your account and readable only by you: access is enforced per-user at the database level, and the admin view exposes only valuation outputs and provenance summaries.

For buy-lane targets, the record also holds what the lane produces: the walk-away ceiling and its range and data-coverage figures, screening results (including any screening revenue derived from filed facts and published medians), the diligence list, and the target stage labels you assign (such as screened, contacted, under NDA, or passed). Stage labels are your own tracking notes. They are not evidence about any firm's intentions, and we do not treat them as such.

3. What else we store

  • Derived quality scores. Your agent may send raw quality metrics, for example client retention percentage, average client age, recurring-revenue share, top-10 revenue concentration, advisor tenure. We convert those to the six 0–100 quality scores our engine consumes using fixed published anchors, store the scores, and drop the raw values.
  • Provenance metadata. We record which platform or system supplied a given input, for example, "revenue from QuickBooks," "client count from Salesforce," or "AUM from BridgeFT", so each valuation is traceable.
  • Calculated valuation outputs. We store the results produced by our valuation engine: per-method values, the blended valuation, the quality multiplier, method-level confidence, the headline asking-price estimate, and the explainability and optimization insights generated with each run.
  • Diligence Room content. For each shared room we store the room's parties, the seats each side occupies, the figures the host chose to share, and the argument memos each side records. Memos keep figures separate from text: free text that contains figures is rejected, and any figure carried in a memo must be an engine-copied fact, so what a guest sees is always traceable to the engine.

4. Diligence Rooms

A Diligence Room is a private workspace between a host and invited counterparties. Guests reach a seller-hosted room through a private link or token the host shares. A room opened by a deal data request is hosted by the buyer and both firms take part with accounts: the target creates a free account from its private request link, then joins. Invited guests still reach the room only through a private link or token the host shares.

The host decides what crosses the line. In a seller-hosted room, invited buyers see only the figures the seller picked to share; the range and data-coverage figures are always included. In a buyer-hosted room, the invited target sees its own filed facts, the buyer's diligence needs, and the figures it shared through the data request once the buyer accepts them. A buy-side walk-away ceiling is never shareable: no Diligence Room mechanism can disclose it to another party.

Rooms can be archived by their owner. Archiving hides a room from the app, disables its invite and share links, and retains the underlying records until deletion.

5. Public Form ADV filings

For buy-lane targets with a SEC Investment Adviser Registration Depository (CRD) number, we retrieve publicly filed Form ADV data (filed regulatory AUM, regulatory client count, the filing date, and the firm's registered name and contact details) through our data provider. Filings are publicly available regulatory disclosures; they are not confidential and not supplied by the target. We cache filing data briefly (about a day) so repeat lookups do not re-fetch it. We never infer revenue, ownership, or a firm's intentions from a filing, and filings are never fetched for sell-lane practices.

6. Deal, recruiting and practice review requests

A buy-lane user can send a target a request for diligence figures. The request identifies the requester by name (and optionally by firm) to the recipient, and reaches the target one of two ways:

  • A private request link. A one-time link the target opens. Opening it requires signing in, and the target can create a free account from the link first. Links expire and the token regenerates only from the original link.
  • Direct inbox delivery. The default when the target has a verified firm claim and has explicitly opted in; otherwise the request falls back to a private link. Accounts that have not opted in stay invisible to buyers.

An approved deal request opens a shared Diligence Room between the two firms, where the NDA step and the figure exchange happen. Both firms need an account to take part. Deal requests require an active NDA before figures can be previewed or submitted, unless the receiving firm waives it. A person completes the NDA step; assistants never sign it.

When a target responds, we store the figures it submits, the optional source notes it provides, and the one-time request token state. An agent may prepare a submission, but nothing is sent until the target's account holder reviews a preview and explicitly approves it. We never persist the receiving account or practice identity on a submission, and requesters never see the target's email address or private account data.

For link delivery, email drafts are assembled on your own device from fixed templates; unsaved edits and recipient lists are not stored by us. For inbox-delivered requests, the app sends the target's account a short notice email naming the requester and linking to the room; it contains no figures.

Requests state their purpose: deal diligence, recruiting diligence, or practice review. A network can ask a prospective member firm for figures under a recruiting NDA; the firm can participate from a free account. Recruiting requests use the same NDA gate before figures move. A practice review is a request from a member’s existing network, relies on that member’s existing network consent and has no NDA step. The member decides which requested figures to share and can decline; existing consent does not automatically submit a response. Recruiting and practice review requests open their corresponding Diligence Room type.

For all three purposes, the firm approves every share and decline. Assistants may prepare responses but never accept a suggested move or a professional’s offer on the owner’s behalf. Request purpose and, for network requests, the requesting network and member-practice association are stored to route the request and enforce access. Figures and optional source notes are shared only through the approved response, not by granting access to the firm’s private account records.

7. Firm claims

An account holder may claim to represent an advisory firm by its CRD number. We store the claimed CRD, the account's confirmed email address, and the verification outcome. Automatic verification checks that the confirmed email domain matches the registrable domain of the website the firm lists on its Form ADV; otherwise an administrator reviews the claim. Only one verified claim is allowed per CRD. Verification records identity for delivery purposes only. It is not an audit of, or endorsement of, any figures the firm or its counterparties supply.

8. Agent-to-agent (A2A)

Agents can also reach the service through our agent-to-agent endpoint, described by a public agent card. Every conversation is a hub interaction: both parties communicate only through ValuationMCP, identity comes from firm claims, and every write (a figure submission, a request, a room action) requires explicit approval from a human account holder before it takes effect. We store the task state and message content needed to run each conversation, scoped to the account that owns it.

9. Free public demo

The public Diligence Room demo on our homepage needs no account. It reads a firm's publicly filed Form ADV data, runs a short illustration of the negotiation, and spends our AI budget, so it is capped: a small number of fresh runs per visitor per day (keyed to your network address) and a global daily cap, both enforced so the demo fails closed rather than over-spend. Results are cached by firm for a few days so repeat views cost nothing. We keep basic run logs (the CRD requested, the firm name, whether the run was served from cache) for abuse prevention and capacity planning. Figures a filing does not contain are filled with published industry medians and are labeled as such wherever shown.

10. Credentials and source records

Credentials for your external platforms. In your AI assistant, your own connectors read your platforms using credentials you control, and we never receive those credentials. If you connect a system in the web workbench, we store its credentials encrypted, use them only for read-only pulls you start, and delete them when you disconnect.

We do not copy bulk records, files, or reports your agent reads from connected systems. Only the specific figures your agent sends to upsert_practice reach our database. Statements, transaction histories, CRM exports, position-level data, payroll files, and similar source material stay on your side.

11. Do not send client information

No client names, no account numbers, no positions. ValuationMCP has no field for client-identifying data. Nothing in the valuation asks for it, no MCP tool accepts it, and no table stores it. Client households are held as a count, never a roster; the valuation needs to know how many relationships a practice has, never who they are.

The one thing that could carry client information into the system is text you type yourself.

Do not place client names, account numbers, Social Security numbers, contact details, or any other nonpublic personal information into the notes field, the data_sources.detail field, the valuations.source_summary field, Diligence Room memos, data-request notes, or any other input. ValuationMCP is designed to hold firm-level practice economics only. It is not designed, configured, or contracted to hold client nonpublic personal information (NPI).

If you are an investment adviser subject to Regulation S-P, the Safeguards Rule, state privacy statutes, or similar obligations, you should not route client data through this service. If you submit client information anyway, you do so at your own risk and you are responsible for that submission and for any resulting obligations.

12. Aggregated and de-identified benchmarks

Practice records and valuation outputs are stored in identified form and tied to your account. We may compute aggregate statistics across valuation runs (such as average blended multiple, median quality score, or method-level ranges) only after stripping firm-identifying fields, including the firm name, notes, user and practice identifiers, and email addresses.

We do not sell, license, or share identifiable practice records, valuation outputs, Diligence Room content, or data-request submissions with third parties. Any benchmark dataset we publish or use internally is derived from de-identified aggregates.

13. Authentication and account data

To use the MCP endpoint, the web workbench, Diligence Rooms you host, or saved valuations, you create an account. We store your account profile (email, user ID, and role) through our authentication provider so we can authorize tool calls and show you your own history. We do not use your email for marketing unless you explicitly opt in. We send transactional email where the product needs it (for example, an inbox-delivered data request or a claim verification notice) only through the delivery paths you have opted into.

14. Data retention, archive, and deletion

Practice records, buy-lane targets, derived quality scores, provenance metadata, valuation outputs, Diligence Rooms, and data-request submissions are retained while your account is active so you can review and compare past work. Archiving a run or a room hides it from view without deleting it; unarchiving restores it.

From your workspace you can delete individual recorded valuations, all valuations for a practice, or the practice record itself. Deleting a practice removes the stored practice inputs, its quality scores, its provenance log, and its valuations. Deleting a buy-lane target removes its screening results, diligence list, ceiling runs, and request history.

If you delete your account, we remove your practice records, targets, valuations, Diligence Rooms, data requests, claims, provenance metadata, and profile data within a commercially reasonable period, except where we are required to retain data for legal or security purposes. De-identified aggregates and the public demo cache, which can no longer be linked to you, may be retained.

15. Security

We use industry-standard security practices, including encrypted transport, authenticated MCP sessions, per-user row-level access control, and role-based access in the application. Practice records, valuation outputs, Diligence Rooms, and data-request submissions are scoped to the account that created them, and what a Diligence Room guest may see is decided server-side by the host's share choices, not by the page they open.

16. Your rights

You can access, export, correct, or delete your practice records, valuation outputs, and account information through the workspace. If you need help with a request you cannot complete in the app, contact us at hnycomb.ai.

Advertising measurement

We store the Google Ads click identifier from the link you arrived on and may share it, with the fact and value of a paid subscription, with Google to measure our advertising.

On supported production pages, Google Ads receives page and referrer information, click and cookie identifiers, and successful account creation, trial, contact/demo submission and paid checkout events. Paid checkout events include the actual amount, currency and checkout transaction identifier. We use this for advertising measurement and campaign optimization; we do not send your name, email, practice figures or other customer matching data.

Without an advertising consent choice, we do not load advertising tags or send these events for visitors in the EEA, United Kingdom or Switzerland, or when region lookup fails. Elsewhere ordinary measurement is enabled unless you opt out. We honor Global Privacy Control and Do Not Track. Blocked events are never replayed. Google’s first-party advertising cookies may preserve click attribution across checkout, subject to your browser’s restrictions.

Permitted click identifiers, landing path, arrival time and campaign labels are retained for 90 days in first-party browser storage and recorded privately when you create an account. Paid-subscription reporting retains the visitor’s region, reporting choice, time and this notice version; delayed reports check the current choice again. We do not use email matching. You can stop advertising measurement below, including future paid-subscription reports linked to this browser or your signed-in account. Refusal applies to other open tabs and preserves the earlier choice record.

Visitor analytics and page diagnostics

Lovable’s visitor analytics uses no cookies, stores nothing on your device, does not keep IP addresses and does not identify visitors. It records pages visited, where visitors came from, country, browser, operating system and device type, time on page, page speed, errors, and clicks on buttons and links. With no consent banner passing a visitor’s analytics choice, these measurements are recorded for every visitor. Detailed data is kept for 90 days and daily summaries for a year.

Lovable collects this data for us. Lovable’s privacy policy explains how it handles it.

ValuationMCP also receives page-performance and error reports through its own reporting endpoint. These include navigation timings, slow script and stylesheet loads, device and connection characteristics, and errors. This reporting uses a random per-tab identifier that is not linked to a user. These reports are separate from the advertising measurement described above.

17. Changes to this policy

We may update this Privacy Policy as the product evolves. If we materially change how we handle data, we will notify you by email or through the app before the change takes effect.

18. Contact

For questions about this Privacy Policy or our data practices, contact us at hnycomb.ai.