Outcome-led recipes. Each one states the real-world problem, the PEAC pieces you'll use, and a step-by-step path from a clean clone to a verified record.
| Recipe | Outcome | Audience |
|---|---|---|
| Runtime evidence export | Export portable evidence from a managed agent runtime for compliance and audit. | Runtime / platform operator |
| API record issuance | Issue signed records on every HTTP response with Express middleware. | API provider |
| MCP tool-call records | Attach signed records to MCP tool-call responses. | MCP server operator |
| MCP gateway records | Issue signed records for gateway-mediated MCP tool calls: policy decisions, content digests, tool-definition references. | MCP gateway operator |
| Commerce evidence bundle | Bundle observational evidence across x402, ACP, and paymentauth without synthesizing payment finality. | Commerce evidence operator |
| Regulatory audit trail | Build a portable, signed audit trail for EU AI Act, NIST AI RMF, and ISO 42001 review. | Compliance / audit lead |
| Agent spend attribution | Review observed agent spend over verified records, without synthesizing balances or finality. | Finance / audit reviewer |
| Cloudflare + x402 + PEAC | Compose Cloudflare delivery surfaces and x402 PR-1986 terms with PEAC policy and terms binding. | Cloudflare-fronted operator |
| Runtime governance composition | Compose a runtime governance toolkit with portable signed records any external party can verify offline. | Platform operator |
| Verify at the edge | Verify PEAC records inside a Fetch-compatible edge runtime with bounded body, key, timeout, and cache behavior. | Edge runtime operator |
| Record evaluation-platform events | Emit portable signed records for evaluation, experiment, and result-artifact events reported by an external evaluation platform. | Evaluation-platform operator |
| Issue harness execution records | Emit portable signed records for each bounded harness run using the existing CLI execution and agent-action surfaces. | Harness / CI operator |
Each recipe follows the same structure:
- The problem — what real situation this addresses, in plain language.
- What you'll use — packages, optional adjacent systems, prerequisites.
- Step-by-step — numbered commands and code.
- Evidence of output — what the record or report looks like.
- Where to go from here — deeper specs, compatibility rows, integrator kits.
Every recipe preserves PEAC's boundary: PEAC carries signed records of what another system attested. PEAC does not execute, orchestrate, evaluate, approve, enforce, or determine payment finality.