exitvalue-mcp

An MCP (Model Context Protocol) server that answers "what multiple do businesses in industry X sell for?" with real M&A data: median, 25th and 75th percentile EV/EBITDA and EV/Revenue multiples, by industry and deal size, from disclosed transactions (2018-2026).

• Free, open source (MIT), no API key, no telemetry.
• Data: the open ExitValue.ai multiples dataset (CC-BY 4.0): by deal size and by deal year, each fetched once and cached in memory for an hour.
• Honest by design: every figure comes with its deal count (n); cells with fewer than the dataset's min_cell_n (10) disclosed deals are withheld, never extrapolated.
• Works with Claude Desktop, Claude Code, Cursor, Windsurf, VS Code and any other MCP client (stdio transport).

Don't want to install anything? Use the hosted connector: https://exitvalue.ai/api/mcp (Streamable HTTP, no auth). Setup guide: https://exitvalue.ai/ai-connector. It is also in the official MCP Registry as ai.exitvalue/ma-multiples.

Tools

All tools are read-only (readOnlyHint, idempotentHint), take no credentials and return plain text ending with a link to the free calculator.

Tool	When to use it	Inputs	Output
list_industries	Find a valid id	none	Industries and sub-vertical ids (dental practice, HVAC, SaaS, trucking, ...), * marks those with published multiples
get_valuation_multiples	"What multiple does X sell for?" (one industry)	industry (id or plain English: "dentist", "med spa", "plumber", "MSP"); optional size_bracket (under_5m_ev, 5m_25m_ev, 25m_100m_ev, 100m_500m_ev, over_500m_ev)	Median + interquartile range with n per enterprise-value bracket
compare_industries	"Which sells for more, X or Y?"	industries (2-6 ids/names); optional size_bracket; optional metric (ev_ebitda default, ev_revenue)	One row per industry: median, IQR and n per bracket; rows below the threshold are marked INSUFFICIENT DATA
get_multiples_trend	"Are multiples rising or falling for X?"	industry; optional metric	Median, IQR and n per deal year 2018-2026 from multiples-by-year.json; years under n=10 are listed as gaps, never interpolated

Example

Question: "What multiple do dental practices sell for?"

get_valuation_multiples({"industry": "dental practice"}) (real output, dataset of 2026-10-29):

Dental Practice - M&A valuation multiples (ExitValue.ai)
By deal size (enterprise value), disclosed deals 2018-2026:
- $5M-$25M EV: EV/EBITDA median 7.0x (middle 50%: 6.5x-8.3x), n=10
- $100M-$500M EV: EV/EBITDA median 11.9x (middle 50%: 11.2x-14.3x), n=14
- $500M and up EV: EV/EBITDA median 14.0x (middle 50%: 11.9x-15.8x), n=17; EV/Revenue median 2.65x (middle 50%: 2.11x-6.18x), n=10
These are population medians; a specific business's multiple depends on its margins, growth, customer concentration and owner dependence.
Source: ExitValue.ai analysis of disclosed M&A transactions (2018-2026) from a licensed historical M&A database, SEC EDGAR filings and verified press releases. Data CC-BY 4.0: https://exitvalue.ai/data/multiples.json

Get a free, business-specific valuation range (adjusted for your margins, growth and owner dependence) in about 2 minutes: https://exitvalue.ai/valuate?src=mcp-oss&industry=dental-practice

Plain-English names are mapped to the closest category and say so (e.g. "plumber" uses HVAC, "MSP" uses IT services, "med spa" uses other healthcare). When a category has no cell with enough deals, the tool says that rather than guessing.

Install

uvx (no install step)

uvx --from git+https://github.com/nickcals/exitvalue-mcp exitvalue-mcp

pip

pip install git+https://github.com/nickcals/exitvalue-mcp
exitvalue-mcp

Requires Python 3.10+. Works with the mcp Python SDK 1.x and 2.x.

Client configuration

Claude Desktop

claude_desktop_config.json (Settings -> Developer -> Edit Config):

{
  "mcpServers": {
    "exitvalue": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/nickcals/exitvalue-mcp", "exitvalue-mcp"]
    }
  }
}

Cursor

~/.cursor/mcp.json (or .cursor/mcp.json in a project):

{
  "mcpServers": {
    "exitvalue": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/nickcals/exitvalue-mcp", "exitvalue-mcp"]
    }
  }
}

Claude Code

claude mcp add exitvalue -- uvx --from git+https://github.com/nickcals/exitvalue-mcp exitvalue-mcp

Zero-install: hosted server

Any client that supports remote MCP servers can skip the local install and point at the hosted endpoint:

{
  "mcpServers": {
    "exitvalue": { "url": "https://exitvalue.ai/api/mcp" }
  }
}

In Claude (web/desktop) add it under Settings -> Connectors -> Add custom connector with the URL https://exitvalue.ai/api/mcp. Step-by-step instructions for ChatGPT, Claude, Cursor and others: https://exitvalue.ai/ai-connector.

Development

git clone https://github.com/nickcals/exitvalue-mcp && cd exitvalue-mcp
python -m venv .venv && . .venv/bin/activate
pip install -e ".[test]"
pytest
npx @modelcontextprotocol/inspector exitvalue-mcp   # optional: poke at it in the MCP Inspector

EXITVALUE_DATA_URL overrides the dataset URL (e.g. to point at a mirror).

Data and license

• Code: MIT (see LICENSE).
• Data: exitvalue.ai/data/multiples.json, CC-BY 4.0. Aggregate medians/quartiles of disclosed M&A multiples by sub-vertical and enterprise-value bracket, built from a licensed historical M&A database, SEC EDGAR filings and verified press releases; cells need n >= 10 per metric and outliers are excluded. Please cite https://exitvalue.ai if you use it. Also mirrored, with a datapackage and citation file, at github.com/nickcals/multiples.

These are population medians, not a valuation of any particular business. For a business-specific range (adjusted for margins, growth, customer concentration and owner dependence), try the free calculator at https://exitvalue.ai/valuate?src=mcp-oss.