Documentation

Build on the record.

Everything an engineer needs to put an agent on a tamper-evident record: the SDKs, the raw HTTP API, and how the proof actually works.

Start hereSDK guide — instrument an agentTwo lines to a tamper-evident record: record(), the human-approval gate, and closing a run. TypeScript and Python.ReferenceHTTP API referenceIngestion, approval gates, run lifecycle, OTLP (protobuf + JSON), evidence packages, and signed webhooks.Deep diveArchitecture — how the record worksThe hash chain, external RFC 3161 anchoring, the oversight path, tenancy, and how everything is tested.The argumentEvidence, not telemetry — whitepaperWhy AI agents need a system of record and observability can't be one: the chain construction, external anchoring, oversight as a runtime primitive, and what we deliberately don't claim.