A team can agree on standards in AGENTS.md or CLAUDE.md in an afternoon. Knowing they’re actually followed — across many people and hundreds of sessions — is the hard part. Here’s how to see it.
Standards everyone agreed to, and no way to see if they held.
The rules are shared — don’t push to main unasked, always run the tests, update the changelog, never commit a secret. But each agent session happens on one developer’s machine, and nothing rolls up “which rules are getting broken, and where.”
Spot-checking one session tells you about one session. For a team you want the pattern: the rule that keeps getting crossed, the repo where it happens most — the thing a code-review norm can then target.
The same transcript read, now multiplied by everyone.
Every developer’s sessions sit in their own home directory:
~/.claude/projects/<project-path>/<session-id>.jsonl
One JSON object per line, per session, per person. AGENTS.md is read the same way — it’s the cross-tool equivalent of CLAUDE.md.
You could open each file and check each rule against each session the way you’d check one. For a team that’s hundreds of files across machines — it doesn’t scale, and nobody keeps it up. What a team needs is the per-session check made repeatable, and the results pooled.
Per-session checks, rolled into one local report — free.
Each developer runs the same one-command check, and can save a small export of the result:
$ npx rulereceipt@latest check --export .rulereceipt/exports # later, pool everyone's exports into one report $ npx rulereceipt@latest team .rulereceipt/exports wrote team-report.html — rules broken most, by whom
Sample. team merges the export files from check --export into one HTML report. It runs locally — nothing is uploaded, no server, no account.
Because check exits non-zero when a rule is broken, teams also run it in CI, so a broken rule can fail a pull request the same way a failing test does. The export + team report then shows the pattern over the sessions you’ve collected.
Honest about the ceiling, same as everywhere: this detects and reports, after the fact. It doesn’t make anyone’s agent comply, and the free team report is local — you pool the exports yourself.
The CLI — check, the per-session report, check --export, and the local team roll-up — is free and source-available. A hosted team tier (trends across people and repos, kept in one place) is how the project is funded and is still being built; there’s nothing to buy yet. See pricing for where that line sits.
RuleReceipt is a free, source-available CLI that checks Claude Code sessions. It also reads AGENTS.md, and the rules files used by Cursor, Copilot and Windsurf. Start with npx rulereceipt@latest check — no account, no upload.