command-line tool

Install. Run. Read.

stackpeek is a Python package. The same package that generates this leaderboard runs on your machine. It is MIT-licensed and has a rules-only mode that works with zero API calls.

install

pip install stackpeek

Or from source:

git clone https://github.com/BladonGarlanddev/stackpeek.git
cd stackpeek
pip install -e .

audit a single site

stackpeek audit https://your-site.example

Emits a Markdown report to standard output. Add -o report.md to write it to disk, or --json to emit structured JSON for your own pipeline.

batch audit

printf 'https://one.example\nhttps://two.example\n' | \
  stackpeek batch -o out.json --concurrency 3

Runs the audit against a list of URLs in parallel, emits a single JSON with one record per site, and prints an aggregate table of vendors that appeared across the batch.

rules-only mode

stackpeek audit https://your-site.example --no-llm

Runs every check that does not require an API call. Produces a report with deterministic findings only — no nuance findings, no LLM-read policy claims. Works with no ANTHROPIC_API_KEY set.

python api

from stackpeek import audit

report = audit("https://your-site.example")

for f in report.mismatches():
    print(f"MISMATCH: {f.title} — {f.detail}")

what it does not do

stackpeek reads the HTML a server returns to a plain browser GET. It is not a headless browser. JS-rendered sites that inject trackers via post-load JavaScript will look sparser than they are. A Playwright fetcher is on the roadmap to close that gap.

It also cannot see CNAME-masked third parties or analyze actual network traffic. Every known limitation is documented in the repository README.

github.com/BladonGarlanddev/stackpeek →