RuleReceipt checks a coding agent's session against your own rules — AGENTS.md, CLAUDE.md, whatever you wrote — and shows precisely what was followed, what wasn't, and why. Starting with Claude Code, more agents next.
“CLAUDE.md is a wish list, not a contract.”
I built this after hitting the same wall myself — running Claude Code daily, writing rules into CLAUDE.md, and having no way to tell whether they were actually followed or just quietly skipped. Vibe coding with it is genuinely good. Trusting it to follow my own rules, without checking, wasn't.
The privacy story, stated plainly, not buried in a policy page.
Runs locally. The only network call is your own Claude API key talking to Anthropic — same as normal Claude Code use.
There is no server of ours in the path. Nothing you check is stored, logged, or seen by us, anywhere.
Read-only access to CLAUDE.md and your session transcript. No hooks installed, ever, without you explicitly asking.
.claude/settings.json that runs automatically on every session. RuleReceipt does none of that. It runs only when you type the command. Nothing automatic, nothing silent.Free while we're early. After that, pay for what you use — no subscription.
Unlimited checks. No card, no signup friction. This is how we find out if it's actually useful.
Buy a pack, use it whenever, nothing recurring. Announced clearly, in advance — never a silent switch.
Straight answers, including where the answer is "not yet."
Not yet. v1 reads Claude Code sessions specifically. AGENTS.md as the rules format is already cross-tool, and support for other agents is the natural next step — but it isn't built yet, so it isn't claimed here.
For the rules that need judgment, yes — the same key your Claude Code setup already uses. Rules that can be checked by plain pattern matching don't need it at all, and cost nothing to run.
Read-only. It reads CLAUDE.md/AGENTS.md and your session transcript, and writes nothing back. No hooks get installed, no settings get modified — not automatically, not ever, without you explicitly asking.
One run of rulereceipt check — regardless of how many rules it checks. Not metered per rule, per token, or per API call.
No — free while we're validating, through a clearly announced date, then credits. If you're already using it before that date, you'll hear about the switch in advance, not find out by hitting a paywall.
Every PASS/FAIL/UNCLEAR comes with a quoted line of evidence specifically so you can check it yourself, not just trust it. It's Claude's judgment on the harder rules — not infallible, which is exactly why the evidence is shown, not hidden.
Yes — same tool, no team plan needed for this. A developer runs it after a session and shares the report directly (a markdown output is built for pasting into a PR or Slack). It's how an engineering lead sees whether the team's agent-assisted work actually followed standards, without a new dashboard to roll out.