ApexClaw Receipts
ApexClaw Receipts is the draft, openly published receipt format for the evidence our governance audit produces — eight objects and seven invariants describing what an agent was permitted to do, what it did, and what it was refused. It is not a standalone protocol, not a ratified standard, and we do not certify anyone against it.
Get an Agent Trust Gap BriefWhat it is, stated plainly
Status: DRAFT. ApexClaw Receipts is a receipt format used inside ApexClaw's own governance audit — not a standalone protocol, and not a ratified standard. It must not be cited as either. No certification exists for it and ApexClaw does not issue one. It is published because a buyer or auditor evaluating our governance audit should be able to inspect the exact shape of the evidence it produces, rather than take our word for it.
The schema is available at /.well-known/apexclaw-receipts.schema.json with worked samples alongside it.
The eight objects
| Object | What it records |
|---|---|
| agent_passport | Identity, owner, permitted action classes, validity window |
| policy_decision | Which policy was evaluated, against which inputs, with what outcome |
| approval | A human authorisation bound to a specific action, signed |
| action_intent | What the agent proposed to do, before it did it |
| execution_receipt | What was actually done and what the provider returned |
| refusal_receipt | What was attempted and blocked, and under which rule |
| revocation | Withdrawal of authority, with effective time |
| replay_job | A re-execution of a recorded decision for verification |
The seven invariants
These are the properties that make the objects useful. A record set that satisfies the schema but violates the invariants is well-formed and worthless.
- Every execution receipt references a policy decision. An action with no recorded authorisation is unexplained by construction.
- Every approval binds to exactly one action intent. Blanket approvals are not approvals; they are permissions with better branding.
- Refusals are recorded with the same weight as executions. Evidence that only records success cannot demonstrate a control.
- Revocation has an effective time, and receipts after it must be refusals. This is what makes revocation testable rather than asserted.
- Records are append-only and integrity-protected. Mutable evidence is not evidence.
- The signing algorithm is a named field, not an assumption. Crypto agility built into the schema rather than bolted on later.
- Replay reproduces the decision, not the side effect. Verification must never re-fire the action.
Why publish it
This is the receipt format our own governance audit produces — not an attempt to set an industry standard. We publish the schema itself, rather than just describing it, so a buyer or auditor can check the exact shape of the evidence against the real file instead of trusting a summary. If it happens to be useful to other teams facing the same evidence problem, implementing or forking it costs them nothing under the licence below.
It is CC-BY 4.0 licensed, expressed as a real JSON Schema document, referenced from page markup as JSON-LD per the schema.org Organization vocabulary. Implement it, fork it, or tell us where it is wrong — that last one is genuinely the most useful.
Why evidence rather than assertion: Gartner projects more than 40% of agentic AI projects will be cancelled by 2027, and EU AI Act Art.99 penalties run up to 7%, 3% or 1% of global turnover depending on the obligation breached — see the breakdown. This is the same evidence format behind Omega, the system ApexClaw governs itself with: 0% real external sends across its entire history, verified on the governed-run record, not asserted.
Common questions
Is this a standalone protocol or a standard?
No. It is the receipt format ApexClaw uses inside its own governance audit — not a standalone protocol, and not ratified by any standards body. It should not be cited as either.
Can we get ApexClaw Receipts certified?
No. No certification exists and ApexClaw does not issue one. Any offer of certification against this schema is not legitimate.
What licence is it under?
CC-BY. Use it, modify it, build on it, including commercially.
Why publish the receipt format you use internally?
Because a buyer or auditor evaluating our governance audit should be able to see the exact shape of the evidence it produces, not take our word for it. Publishing it is self-interested: an auditable claim is worth more than an asserted one.
How does it relate to OWASP or NIST frameworks?
It is complementary and narrower. Those frameworks describe risks and management practices. This schema describes the shape of the evidence our governance audit produces when you implement controls against them.
A receipt format anyone can inspect is worth more than a proprietary one, because trust that cannot be verified independently is just branding.