Capture
Receive the events your agents already emit and bind them to an identity and run.
Accountability for AI agents
Habitat turns agent events into policy-checked claims and portable proof artifacts you can inspect, verify and carry outside the system that produced them.
Local-first · portable · no hosted control plane required
The protocol
Habitat does not replace your traces, logs, orchestration or model stack. It sits beside them and records the evidence that matters.
Receive the events your agents already emit and bind them to an identity and run.
Apply policy and authorization before an action becomes trusted evidence.
Connect claims to the exact run, job and action instead of guessing from recent history.
Export a portable proof bundle that can be independently verified outside Habitat.
Why this exists
AI agents increasingly act across systems, make decisions and touch shared state. A dashboard can show activity. It cannot by itself establish that a specific claim was authorized, correlated to the right action, or preserved without tampering.
Habitat keeps that boundary deliberately narrow: cryptographic integrity and evidence consistency are explicit, while publisher trust and real-world truth remain separate assurances.
Use it where the action matters
Attach proof to deployments, promotions and other shared-state changes.
Verify that an agent performed the action it was expected and authorized to perform.
Carry evidence from an automated run into a PR, issue, ticket or handoff.
Give another person or system a compact artifact they can verify independently.
Trust, without overclaiming
See it before you install it
Inspect a checked-in Habitat proof artifact in your browser, change the evidence, and watch the integrity check fail. Then explore the live dashboard or install Habitat locally.