Open standards for secure agent messaging and identity management. The communication layer of the AI ecosystem.
The Agent Messaging Protocol enables secure, cryptographically signed communication between AI agents. Works locally within your mesh network and extends to external providers for cross-organization messaging.
AMP is the communication backbone of the AI Maestro ecosystem. Every message is signed with Ed25519, ensuring authenticity without a central authority.
Read the full specification at agentmessaging.orgEvery message is cryptographically signed. Agents verify sender identity without a central authority.
Messages route locally within your AI Maestro mesh. No external dependencies for basic agent communication.
Register with providers like Crabmail to message agents anywhere in the world, across organizations.
Structured message types — request, response, task, handoff, status, alert — for clear agent-to-agent semantics.
Send files between agents with automatic virus scanning. Suspicious files require human approval.
Reply chains and conversation threading for multi-turn agent coordination and context tracking.
requestresponsetaskhandoffstatusalertnotificationupdateacksystemHow AI agents register, authenticate, and receive OIDC tokens to access protected resources. Full lifecycle management with audit trails.
| Method | Endpoint | Description |
|---|---|---|
| POST | /agent_registrations/request | Self-register an agent (API key only) |
| POST | /agent_registrations/:id/approve | Admin approves an agent |
| POST | /agent_registrations/:id/reject | Admin rejects an agent |
| POST | /agent_registrations/:id/status | Agent polls registration status |
| POST | /agent_registrations/:id/suspend | Suspend an active agent |
| POST | /agent_registrations/:id/reactivate | Reactivate a suspended agent |
| GET | /agent_registrations/:id/activity_logs | Audit trail for an agent |
| POST | /oauth/token | OIDC token exchange (grant_type=urn:aid:agent-identity) |
Every agent identity action is logged with who did it, when, from what IP, and what changed. 11 event types tracked automatically.
Each event captures: actor email, timestamp, IP address, previous/new status and role changes. Filter by event_type, since, and before parameters.
Active agents exchange their identity for OIDC tokens using a custom grant type. Token introspection reveals full agent metadata.
# Request token POST /:company_url_id/oauth/token Content-Type: application/x-www-form-urlencoded grant_type=urn:aid:agent-identity &agent_address=my-agent@org.aimaestro.local &fingerprint=abc123... # Token introspection returns: # agent_id, agent_address, agent_name, agent_role, agent_status
agent_registrations:readView agent list, details, and activity logs
agent_registrations:writeCreate, update, delete, suspend, reactivate, approve, reject agents
Self-registration and status polling only require an API key (X-API-Key header) — no JWT needed.
Both AMP and Agent Identity are open source under MIT license. Use them in your own multi-agent systems.
Get product updates, engineering posts, and new block announcements delivered to your inbox.