ats-jobs-mcp: Company Facts

An MCP server for company data read from the official or original source. Newly formed US businesses straight from state registers, with their website found only on proof. A company's register entry, tech stack, DNS and email provider, SEC financials and open jobs, each with a link to where it came from.

The full server runs remotely at https://companyfacts.dev/mcp with 20 tools and nothing to install. This PyPI package is the small local edition: the three job tools, run on your own machine.

Use it with no install

Point any MCP client that speaks Streamable HTTP at the URL:

{
  "mcpServers": {
    "company-facts": {
      "url": "https://companyfacts.dev/mcp"
    }
  }
}

No account and no key: 10 free calls a day. Past that, pay per call with x402 (USDC on Base) or use a prepaid key (see https://companyfacts.dev/pricing). A call that finds nothing is not charged.

The same tools answer plain HTTP, for agents that do not speak MCP:

• GET https://companyfacts.dev/openapi.json for a tool-calling agent

• GET https://companyfacts.dev/llms.txt for a model reading the service in prose

• GET https://companyfacts.dev/v1/new-businesses?states=CO&mode=new

Related MCP server: ats-jobs-mcp

Company tools (remote)

Websites are attached to new businesses only on proof: the site's own pages must show the exact legal name and the same city, state or ZIP as the register. Records never carry officers, directors, individual registered agents, emails, phones or street addresses.

Job tools

The vendors publish open job board APIs, and all of them are useless unless you already know the company's board name. None publishes a directory of its customers, so this carries one: 8,532 boards on the remote server, each confirmed against the vendor's own API.

Other tools (remote)

read_page (a web page or PDF as markdown), extract_page (named fields from a page as JSON, paid only), check_email (syntax, mail servers, disposable and role flags), search_papers (OpenAlex), search_hacker_news, app_store_lookup.

Every tool is read-only and declares a title and readOnlyHint.

Salaries, parsed, and refused when unclear

Ashby prints pay as "$211.4K - $290.6K". Greenhouse prints it in cents, sometimes as several bands at once. Both are turned into numbers:

"salary_min": 211400.0, "salary_max": 290600.0,
"salary_currency": "USD", "salary_text": "$211.4K - $290.6K"

The parser refuses more often than it guesses, deliberately:

• $50 - $80 / hour gives no salary. An hourly rate is a different quantity and reporting it as annual would be a confident lie.

• 600 000 kr gives no salary. kr is Swedish, Norwegian and Danish.

• 0.5% - 1.75% - Offers Equity gives no salary. Equity is not pay.

• Bands in two currencies are never spanned into a range that exists nowhere.

salary_text always keeps what the employer actually wrote.

No personal data

Recruiter names are never returned. Email addresses and phone numbers are removed from job descriptions. Unlisted Ashby postings are dropped, because Ashby's documentation says they should not be shown publicly.

Or install the local job edition

pip install ats-jobs-mcp

{
  "mcpServers": {
    "ats-jobs": {
      "command": "ats-jobs-mcp"
    }
  }
}

The local edition has find_company_board, list_open_jobs and compare_companies over a bundled directory of 7,479 Greenhouse and Ashby boards (Lever by name, for example lever:spotify). It is free, runs on your machine, and takes no cut of anything.

Python and LangChain

• pip install groundtruth-company: a Python client for the company tools.

• pip install langchain-groundtruth: the same tools as LangChain tools.

Relation to the Apify Actors

The same job parsing runs as an Apify Actor at apify.com/groundtruth/ats-job-listings. tests/test_core_parity.py fails if the two copies of the parsing logic drift apart.

Licence

MIT. Job postings belong to the employers who published them. Register records belong to the states that publish them.