Open standards and infrastructure for trustworthy AI agents.
Gazebo builds the trust layer for agents: portable identity, consent-aware access, runtime policy, and auditability.
- Agent Identity Protocol (AIP) — the public specification, schemas, shared fixtures, and TypeScript/Python validators.
- AIP specification — read the published v0.1 draft.
- Identity schema
- Policy schema
- Audit log schema
- Gazebo — agent infrastructure and reference implementation.
The Agent Identity Protocol is one open standard with three independently adoptable sections:
- Identity — declare who an agent is and what it needs.
- Consent — make user authorization explicit and verifiable.
- Runtime — enforce policy and record what happened.
AIP is designed to be portable across agents, tools, providers, and runtimes. The canonical schemas and conformance fixtures are public so implementations can validate against the same contracts.