Work Items
Active specifications and protocols under development.
Agent Identity & Discovery (AID)
A lightweight DNS-based protocol that lets any client discover an agent's endpoint, capabilities, and config with a single query. AID is the foundation for agent discoverability on the open web.
Status: v1 specification published, SDKs in development.
AID Spec & Workbench
Full specification, interactive workbench, and aid.json generator.
SDKs & Tools
Open source SDKs in TypeScript, Go, Python, and more.
What AID solves
| Problem | AID solution |
|---|---|
| "Where is the API?" | DNS TXT record points to agent config |
| "What can this agent do?" | Structured capability declaration in aid.json |
| "Is this agent legitimate?" | Tied to domain ownership — DNS proves control |
| "How do I call it?" | Endpoint, auth method, and protocol in one place |
Get involved
- Review the spec and file issues on GitHub
- Join the Discovery working group on Discord
- Implement AID for your agents and share feedback
Future work items
The community is scoping additional specifications. Areas under discussion:
- Agent verification — trust signals and reputation beyond DNS
- Inter-agent protocols — standards for agent-to-agent communication
- Safety frameworks — shared norms for agent behavior and accountability
Want to propose a new work item? Start a discussion on GitHub.