Edgartools

Find the latest up to date information by visiting https://edgartools.readthedocs.io/en/latest/.

Edgartools is a python wrapper around the SEC EDGAR API. It's extremely useful for navigating anything that is publicly accessible through this API.

This includes but is not limited to:

• Explore companies
• Get latest filings of all types
• Manipulate structured xbrl data
• Download entire filings as text or markdown

Some tips

• You must always set an identity with edgar.set_identity("First Last first.last@example.com").
• This format of "name email" is enforced by the SEC.
• There are convenience methods to access a company's latest filings, see docs/filings.md
• Do not veer into the deep end of XBRL parsing unless explicitly asked
• Most documents can fit into your context window.
• However, if you must partially read a filing, read the appropriate report.

Reading filings

Essential setup

Install with uv

uv add edgartools

>>> import edgar # Notice the difference between package install name and import name. This is extremely important.
>>> edgar.set_identity("name email@example.com")

Fetching company filings

>>> filing = edgar.get_by_accession_number("0000104169-25-000191") # fetch filing by ID - no need to create Company class.

>>> company = edgar.Company("WMT") # protip: you can use CIK, or ticker here.
>>> filing = company.latest(form=["10-K", "10-Q"]) # protip: you can specify multiple form types here
>>> filing.markdown() # get entire filing as markdown
>>> filing.text() # get entire filing as beautifully rendered text. Prefer this as output for readability.

>>> # You may save the entire file to /tmp/ for fast repeated grepping
>>> with open(f"/tmp/{filing.accession_number}.txt", "w") as fh:
...     fh.write(filing.text())

>>> filing.reports # get list of specific reports (disaggregated revenues, etc)
>>> print(str(filing.reports[37].text())) # get a specific report as text

A full filing may be too large to read at once with the Read tool. However, you can either split the file into 3-4 chunks, or read only the relevant reports directly.