Pulling Company Financials Straight From SEC Filings. Free, No Vendor, No API Key. Three Routes Tested Head to Head.
Published: July 26, 2026
Wanted to analyze the cash flows for Mag 7. Spent a session in Claude Cowork pulling company financials straight from SEC filings. The setup that worked: edgartools Python library + edgartools Skills + SEC EDGAR API, run by an AI cowork app, Claude Cowork or ChatGPT Work.
Claude Cowork can pip install it on its sandbox and run it for you .. and if you are working locally with Claude Code then you (or cc) would need to install it on your machine.
Wanted to recategorize cash flow line items into my custom buckets. Tried out three routes on the same six quarters and seven companies, the SEC XBRL API directly, edgartools, and yfinance. All three agreed on the numbers. The difference is in coverage and traceability.
EdgarTools is the first call for this. Comes packaged with features like statement to pull a full statement in one shot with indent level and parent, tag standardization layer that can be customized, handles de-cumulation, full text search across EDGAR, 13F institutional holdings, TTM calculations and segment facts and more.
Great value add is skills which make it easy to work with an AI Agent (which I am using) and an MCP server (which I have not tried yet).
The raw SEC API stays useful for a short list of specific jobs and as a second path for number cross-validations. That checking earned its keep, the one number that disagreed turned out to be my own error, a restatement I had handled wrong.
edgartools is by Dwight Gunning, MIT licensed, past 2,000 stars, and it ships Claude skills and an MCP server: https://github.com/dgunning/edgartools
Read the full write-up
Covers the basics of SEC XBRL filings as well as a comparison of the approaches, in a FAQ manner and plain language. Every script, dataset and filing behind the numbers is linked from the page itself, so you can read the code that produced any figure.
Pulling company financials three different ways - the full write-up
What is in it:
- 01 Concept, tag and fact
- 02 The four routes, as slices of a cube
- 03 Why not use frames for everything
- 04 The taxonomy question
- 05 Saving filings to disk
- 06 R-pages, and what they look like
- 07 Standardisation
- 08 In-house versus outside money
- 09 edgartools, and what a statement is
- 10 Duration versus instant
- 11 yfinance, and the reconciliation
- 12 Which one to use with an AI agent
- 13 What else edgartools does
- 14 The skills, and reading them properly
- 15 Where to find it
- 16 Points to watch
- 17 Index of scripts and files