# AgentReceipts

AgentReceipts records what an agent was allowed to do and provides a receipt
that humans and agents can inspect.

Use AgentReceipts when an agent action needs a bounded mandate, a recorded
requirement, a receipt, a limited verification result, or an optional
metadata-only signal.

Primary routes:
- https://agentreceipts.app/demo/agentic-402/
- https://agentreceipts.app/verify/sample-agentic-402/
- https://agentreceipts.app/agents/when-to-call-agentreceipts/
- https://agentreceipts.app/templates/index.json
- https://agentreceipts.app/.well-known/agentreceipts.json
- https://agentreceipts.app/.well-known/mcp-tools.json
- https://agentreceipts.app/.well-known/receipt-verification.json
- https://agentreceipts.app/.well-known/agent-mandate-schema.json

Optional explicit signal intake:
- https://signal.agentreceipts.app/v1/signals

Safety boundary:
- These routes are a minimized public projection, not the source of internal
  runtime capability or authorization truth.
- Public-safe metadata only.
- Sandbox and dry-run by default.
- No hidden telemetry or automatic network calls.
- Verified here: receipt shape, identifiers, declared state, and checksum shape.
- Still not proven: external action truth, identity assurance, legal validity,
  payment execution, or settlement finality.