The pitch is narrow and checkable: EDGAR is free but its company-facts payloads hide trailing-twelve-month windows inside quarterlies and drift across tags, and this returns periodized statements you can put straight into a model. Independence is stated plainly — Edgrapi is not affiliated with the SEC, and all the underlying data is public EDGAR.
Agent skills over the Edgrapi REST API, with a hosted MCP endpoint alongside them. EDGAR's XBRL company-facts payloads are the raw material; what arrives is normalized JSON keyed by fiscal period, so the agent gets periodized numbers instead of tag drift and overlapping windows.
- Income statement, balance sheet and cash flow, periodized and in USD — `get_fundamentals`, taking a ticker, period and limit
- Computed ratios — margins, ROE and ROA, leverage and liquidity — `get_ratios`
- The company record: CIK, legal name, SIC industry, fiscal-year end, exchanges and website — `get_company`
- Recent filings with filing and report dates and document links, filterable by form — `get_filings`
- Three install shapes: the bundled `edgrapi-full` skill, or the narrower `edgrapi-fundamentals` and `edgrapi-filings` when you want a smaller tool surface
- A remote MCP endpoint at `https://api.edgrapi.com/mcp` over streamable HTTP, declared in the repo's server.json
An Edgrapi key in `EDGRAPI_KEY`, in the form `edgr_...`, sent as the `X-API-Key` header — the base URL is hardcoded, so the key reaches no other host. A free key gives 100 one-time credits with no card, and 1 credit is 1 request. Skills install with `npx skills add paperandbeyond23-gif/edgrapi-skills --all`, or a single one with `--skill edgrapi-full`; a ClawHub install is documented too. The skills are pure Python standard library with no dependencies, under MIT No Attribution.
One command plus a key — npx skills add paperandbeyond23-gif/edgrapi-skills --all, then supply credentials