What it does:
- captures tool calls (input/output/timing/status) via the OpenClaw hook surface
- appends them to a tamper-evident chain (each entry links to the previous digest)
- emits a signed, portable receipt per tool call (JSON you can verify offline)
Why: If you’re running agents across services, the “truth” of what happened ends up split across logs. This gives you a single artifact you can attach to an incident, audit, customer ticket, or cross-team handoff.
Try:
- git clone https://github.com/peacprotocol/peac - cd peac && pnpm install - cd examples/openclaw-capture && pnpm demo
It prints a trace + emits 3 example receipts.
Looking for feedback from OpenClaw users: - which fields are missing for real debugging/auditing? - do you want this to map to an existing trace model (OTel, etc.)? - what storage/transport would you actually use for receipts in production?