PRODUCT / EVIDENCE
Proof Receipts.
Know what was checked and why.
A proof receipt is the audit primitive produced around an AI action decision: proposal, checks, policy, decision, evidence and time.
Receipt anatomy
{
"agent": "sales-agent-17",
"action": "send_campaign",
"policy_version": "marketing-v3",
"checks": ["claims", "audience", "consent"],
"decision": "WARN",
"score": 81,
"evidence_hash": "sha256:...",
"evaluated_at": "..."
}Why receipts matter
When an agent acts, teams need more than a log saying “the model decided.” A receipt creates an explicit, inspectable artifact connecting the proposed action to the policy and evidence used at the time of decision.
Roadmap direction
Proof receipts can evolve toward signed or tamper-evident records, policy-version pinning and interoperable evidence formats. Those are product directions; live NexusEval already uses deterministic hashes in some evaluation outputs.