Agent Identity & Discovery (AID)

The DNS-based discovery protocol for the agentic web.

Agent Identity & Discovery (AID)

The 0-th hop for agent discovery.

Type a domain. Find the current agent endpoint.

AID is a small, DNS-first standard for the first contact with an agent or tool service.

It answers one question: "Given a domain, where does the agent interaction begin?"

One _agent TXT record maps a domain to an endpoint, protocol hint, and optional endpoint-proof key. From there, the selected protocol, auth, policy, and capability layers take over.

Start with Core Concepts for a quick overview, then follow the Quick Start to publish your first agent.

Try it Now — Live Resolver & Generator

Why It Matters

AID gives clients a stable starting point without creating a central registry.

You want to connect your Notion to a new AI assistant. You type notion.so. The connection happens automatically. Your experience is instant.

As an autonomous agent, you're tasked with analyzing a dataset stored in a Supabase project. You don't need a hard-coded endpoint. You can discover the supabase.com agent endpoint, then continue into the protocol and auth layer.

How It Works: The 30-Second Explainer

The entire mechanism is a single DNS lookup. It's simple, decentralized, and built on infrastructure that has powered the internet for decades.

Publish: A provider (e.g., supabase.com) adds one TXT record to their DNS at a standard location: _agent.supabase.com.
Discover: A client, given supabase.com, makes a single DNS query for the TXT record at that address.
Connect: The record contains the agent's uri. The client uses it to connect directly.
Verify: If a public key (k) is present, the client performs PKA endpoint proof before trusting the endpoint. (Optional)

Try this flow now

Rendering diagram...

Learn more

Want the deep dive?

Specification – The exact format, algorithms, and security rules.
Core Concepts – How DNS discovery, protocols, and identity fit together.
Identity & PKA – What endpoint proof adds and where its boundary is.
PKA Endpoint Proof – The exact HTTP signature profile for implementers.
Rationale – The design philosophy behind AID.
Security Best Practices – DNSSEC, redirect handling, local execution, IDN safety, TTL & caching.
Enterprise Rollout – Change windows, delegation patterns, and DNS team vs app team ownership.
aid-doctor CLI – Validate, secure, and generate AID records.

SDKs and Tools

AID has official libraries and tools across multiple languages, with additional ports in progress.

TypeScript/Node & Browser: @agentcommunity/aid
TypeScript Core Library: @agentcommunity/aid-engine – Pure business logic for discovery, validation, and PKA
CLI Tool: @agentcommunity/aid-doctor – Validate, secure, and generate AID records (wraps aid-engine)
Python: aid-discovery (transfer to community planned)
Go: aid-go
Rust: aid-rs
.NET: aid-dotnet
Java: aid-java
Web Workbench: Interactive generator/resolver

See the full package overview in SDKs and Packages and the cross-language Discovery API.

Use Cases

Simple connections: Type a domain, connect automatically. No manual setup.
Stronger trust (optional): Add identity proof to ensure you’re talking to the right service. See Identity & PKA.
Local and dev workflows: Safely run local agents with explicit consent; discover dev agents on your network.
Multi-protocol ecosystems: Connect across MCP, A2A, OpenAPI, gRPC, GraphQL, or WebSocket.
Smooth migrations: Deprecate old endpoints gracefully with clear timelines.