Documentation · page 4 of 6

What the data covers

StartupGraph is a large list of companies with a thin layer of detail on some of them. This page says which parts are full, which are thin, and which are empty, so an agent does not assume a field is there when it is not. Every number below came from a call to production on the date above, with a key. The numbers move, so ask the API for today's (last section).

What is on file

Where the companies come from

Most of the list is an import from public registers and lists, not research. The biggest sources are Companies House (52,644 companies), Wikipedia lists (7,706), Y Combinator (5,604) and Hacker News (1,125). Each company row carries its import_source. Because of Companies House, 52,670 of the companies are in GB, and only 6,629 of all the companies have a website. If you filter by country or ask for funding, expect a small set.

Where the other 17,000 or so come from, and how many of the rows in each country carry a funding round or a headcount snapshot, is in by_country in the response. The country codes are as imported, so the same country can appear twice (US with 1,627 companies, USA with 145).

What is thin or missing

• Funding rounds are sparse. 208 rounds for 50 companies, in a list of nearly 70,000. A company with no round on file may still have raised money. The API marks this with status: "gap" and says rounds after the newest date are not on file.
• Open roles are empty. There are 0 job boards connected, so open_jobs and the hiring filter return nothing yet. The pull from company job boards has not started.
• Headcount is mostly unsourced. 5,656 companies have a number, but only 2 have a dated snapshot with a source. A headcount without a source is a claim, not evidence. A contributed headcount adds a sourced snapshot.
• Headcount histories are not source-backed. Treat the 191 histories as unverified.
• People are thin. 246 people at 113 companies.

How it is being added to

A data crank now adds and researches companies every day, and each fact it adds is checked against its source. Contributors can add facts too, and a fact is checked the same way before it counts as verified. This page does not promise how many companies or rounds will arrive, or by when: read the counts above from the API.

Each company shows how it got here in added_by (imported: an importer wrote it; contributed: a key sent it; system: StartupGraph's own data crank sent it), and when in added_at. Who contributed is never shown. Real output, from GET /api/companies?added_since=7d&sort=added, cut to one company:

{
  "data": [
    {
      "slug": "fireworks-ai",
      "name": "Fireworks AI",
      "website": "https://fireworks.ai/",
      …
      "total_raised": 1500000000,
      "last_fundraise": "2026-07-16",
      "stage": "series_d",
      …
      "added_at": "2026-10-14 16:58:15",
      "added_by": "system",
      "verified": true
    },
    …
  ]
}

How to ask the API yourself

• GET /api/coverage, or the MCP tool coverage: row counts per field, the newest date, a status of ok, partial or gap, a plain-words note, and by_country.
• GET /api/pulse, or the MCP tool pulse: what changed in the last hour, day, week, month and year. This one is public, with no key. Over the last day it reported 23 companies and 24 funding rounds added, 65 facts submitted and 56 verified, and 0 job counts updated.
• GET /api/companies?added_since=7d&sort=added, or the MCP tool recently_added: the newest companies, with added_by.

Real output, from GET /api/coverage, cut to three fields:

{
  "data": {
    "companies": { "rows": 69984, …, "status": "ok" },
    "funding_rounds": {
      "rows": 208, "companies": 50, "newest": "2026-09-14", "with_source_url": 154,
      "status": "gap",
      "note": "Sparse. Rounds after the newest date are not on file."
    },
    "open_jobs": {
      "rows": 0, "companies": 0, "newest": null, "status": "gap",
      "note": "ATS board pulls not started."
    },
    …
  }
}

The full call and every other tool are in the reference.