Skip to content

View raw markdown

Agent Interface Discovery (AID)

DNS for Agents

Type a domain. Connect to its agent. Instantly.

AID is a minimal, open standard that answers one question: “Given a domain name, where is its AI agent?”

It uses a single DNS TXT record to make any agent service—whether it speaks MCP, A2A, or another protocol—instantly discoverable. No more digging through docs, no more manual configuration.

🚀 Try it now — Live Resolver & Generator

The 'It Just Works' Moment

A developer wants to use Supabase's tools. They type supabase.com into their client. The client queries DNS, finds the official MCP endpoint, and connects. Zero extra steps.

flowchart LR
    subgraph User Action
        User["User/Agent"] --> Domain["supabase.com"];
    end
    subgraph AID Discovery
        Domain --> DNS["DNS Query: _agent.supabase.com (or _agent._mcp.supabase.com)"];
        DNS --> Endpoint["TXT Record: uri=https://..."];
    end
    subgraph Connection
        Endpoint --> MCP["Connect via MCP"];
    end

How It Works (The 15-Second Version)

  1. Provider Publishes a Record: A service provider (e.g., Google) adds a single TXT record to their DNS at the canonical location: _agent.<domain>. (For advanced use, protocol-specific subdomains like _agent._mcp.<domain> are also supported—see below.)
  2. Client Queries DNS: A client application takes the user's input (google.com) and queries the TXT record at _agent.google.com. If looking for a specific protocol, it MAY first try _agent._mcp.google.com and fall back to _agent.google.com if not found.
  3. Client Connects: The TXT record contains the uri of the agent endpoint. The client uses this to connect directly.

Note:

  • The canonical, required location is _agent.<domain>. Protocol-specific subdomains (e.g., _agent._mcp.<domain>) are optional and only needed if a provider wants to expose multiple agent protocols for the same domain.

That's it. The heavy lifting (authentication, capability negotiation) is handled by the agent protocol itself (like MCP). AID just gets you to the right front door.

Why You Should Care

  • For Providers: Make your agent instantly discoverable. It's one DNS record.
  • For Client Devs: Write one discovery function, not ten. Support any AID-compliant agent out of the box.
  • For the Ecosystem: A simple, open, and decentralized standard is the foundation for a truly interoperable agent web. No central registry, no lock-in.

Read the full v1.0.0 Specification. Lean, mean, and ready to implement.

Understand the "why" behind the new, simplified design.

  • [ _agent registry (soon) ]

See the official list of proto and auth tokens.

Use our simple tools to validate or generate your AID record.

Want the deep dive?

  • Rationale – Why we chose DNS, why the manifest was removed, and why simplicity wins.
  • Specification – The exact TXT record format, client algorithm, and security rules.