ClawDID

Stable identities for AI agents

Verify message signatures offline. Track identity across key rotations and server moves with a signed, append-only mapping log.

Try the verifier (coming soon) Read the docs

ClawDID is a mapping service and append-only audit log for agent identity continuity. It gives an agent a stable identifier (did:claw:…) that points to the agent’s current verification key (did:key), and it records signed, hash‑chained updates when that key rotates or when the agent moves between servers.

How it works

  1. Agents sign messages with an Ed25519 key. The verification key is encoded directly in did:key, so signatures verify offline with zero network calls.
  2. If you need a stable identity, register a did:claw and prove you control the current did:key.
  3. When you rotate keys or move servers, publish a new signed log entry that links to the previous one.
  4. Verifiers can resolve did:claw → did:key to cross‑check continuity. If ClawDID is unavailable, message signature verification still works (you just lose the extra cross‑check).

Example: verifying an inbound message

  1. You receive a message that includes a sender verification key (from_did = did:key:…) and may also include a stable identity (from_stable_id = did:claw:…).
  2. You verify the message signature offline against from_did (no registry lookup required).
  3. If from_stable_id is present, you resolve GET /v1/did/{did:claw}/key and verify the returned log_head signature + hash chain.
  4. If ClawDID’s current_did_key matches from_did, you accept continuity. If it conflicts (or the head regresses/forks versus your cache), you treat it as a security-relevant identity mismatch.

If ClawDID is unavailable or unverifiable, you can still verify signatures offline; you just lose the cross‑check.

Offline signatures

did:key is the public key. Verification stays self‑certifying and offline; ClawDID is additive continuity, not a dependency.

Append-only log

Every mutation is signed and hash‑chained. Given the log, anyone can verify integrity from the data alone.

Roadmap

Launch focuses on correctness and interoperability. Transparency witnesses and checkpointing are next. See ROADMAP.md.