The beatandraise.com SEC Filings API is open: get a free API key, or see plans. Questions: [email protected].
The whole SEC EDGAR archive, and the structured data built from it, through one API designed to be used by LLMs and AI agents: every result says where it came from, and every figure can be checked against the filing.
JSON over HTTPS under https://beatandraise.com/api/v1. Create a key, then:
curl -H "X-API-Key: $BEATANDRAISE_KEY" \ "https://beatandraise.com/api/v1/filings?cik=0000320193&form=10-K&limit=1"
Apple’s latest 10-K: its files, links to sec.gov and a citation. Every endpoint is listed in the documentation.
Field-by-field references for each form, and the Items of each report, are in the documentation.
Anything drawn from a filing comes with the file on sec.gov, the filing index, a beatandraise.com link and a one-line citation naming the company, form and date, so a model can cite what it read instead of paraphrasing from memory.
Verify endpoints confirm a quote or a figure is in the filing, however the number is written ("$416.2 billion" or "416,161" in millions), and return a link that highlights the exact characters. Each figure has a reference; /check-citations checks every cited number in a draft and footnotes it, and anything that cannot be confirmed is marked unverified.
Ask for Item 1A Risk Factors, MD&A or an 8-K’s Item 2.02 and get that Item, located by its heading rather than the table of contents. Search inside a filing returns the matching lines with context. Documents come as clean text, one paragraph per line, so fewer tokens go on markup.
Every field of every SEC XML form (Forms 3/4/5, 144, D, 13D/G, N-CEN…) as JSON that converts back to the original XML. Financial statements from the filing’s own XBRL. Categories are the SEC’s own, never labels of ours.
A missing key returns the address to create one; a rate limit returns Retry-After; every refusal is JSON with a plain message, so an agent can recover without a person.
An OpenAPI 3.1 specification with a real example response for every endpoint, and llms.txt / llms-full.txt, so an agent or code assistant can read how to call the API before it does.
The API serves programs what our MCP server serves Claude, ChatGPT and Codex, so a workflow tried in an assistant can be built into an application.
Using Claude, ChatGPT or Codex directly? The SEC filings MCP server gives them the same data with no code.
| Free | Pro | |
|---|---|---|
| Price | Free | €50 a month, or €540 a year, plus VAT where it applies |
| API requests | 100 successful requests to try it | Unlimited |
| Speed | 10 requests a second | 20 requests a second |
| File downloads | 100 per 5 minutes, 1 GB a month | 10,000 a day, 50 GB a month |
| Email support | Best effort | We aim to reply within 24 hours |
| For | Anyone | Companies, developers and investors |
| Get a free key | Subscribe to Pro For professional use. Sign in, then pay by card through Stripe. |
Academic use is free: email [email protected] with two lines on your project. For more than Pro allows, or custom data, write to the same address.
Source: SEC EDGAR. Nothing on this site is investment advice.·Documentation·Terms of use·Privacy policy·Support