Announcing AID: identity for the agentic web

TL;DR (ELI5): Type a domain to find its agent, then prove it’s the real one with a quick key-based check.

Simple. Trusted. Works today.

The idea in one sentence

AID is the internet’s address book for agents. Give it a domain, get the right agent endpoint, and optionally verify it with a lightweight cryptographic proof.

What it solves

• No discovery glue: Stop hard-coding endpoints or hunting through docs. One record tells clients where to connect and which protocol to speak.
• Trust at first contact: Optional identity proof confirms the server controls the key the domain published. Clear signal for security teams.
• Works across stacks: MCP, A2A, gRPC, GraphQL, WebSocket. Same discovery, different protocols.
• Pragmatic in real orgs: DNS is ideal. If you cannot touch DNS this week, serve a .well-known/agent file while you line up the change.

Why adopt now

• Faster integrations: Any compliant client can discover and connect to your agent automatically.
• Clear security posture: Turn on key-based proof for admin or high-trust paths. Rotate keys without breaking clients.
• Better UX out of the box: Add a short description, a docs link, and deprecation timelines. Clients can guide users before they connect.

What you can expect

• Minimal surface area: One TXT record. Optional identity.
• No central registry: If you control a domain, you can publish today.
• Stable path forward: The _agent label carries into future versions and tooling.

Learn more

• What AID is and why it exists → AID rationale
• Specification → AID specification
• Identity & PKA, high level → identity / pka docs
• aid-doctor: validate and generate → aid_doctor docs

Join the Agent Community to collaborate, ship examples, and get early access to .agent domains: agentcommunity.org