It's surprising how often the solution to a new, complex problem is a simple, old idea.

Connecting to AI agents right now feels like a step backward—a mess of bespoke configs, hunting for URLs in docs, and writing brittle glue code. For a new layer of the internet to flourish, this can't be the status quo. The agentic web needs its own "dial tone."

AID applies the DNS discovery model, which has worked flawlessly for email for 40 years, to this new problem. The philosophy is simple: use the lowest, most reliable layer possible to solve the most fundamental problem. For service location, that layer is DNS.

Today, we're drawing a line in the sand and shipping AID v1.0 as a finished, stable standard.

The Core Mechanism

The entire specification rests on a single DNS TXT record at a well-known location: _agent.<domain>.

_agent.example.com. 300 IN TXT "v=aid1;p=mcp;uri=https://api.example.com/mcp"

That's it. It's an intentionally minimal primitive.

We deliberately chose TXT for v1 for one reason: universal accessibility. The theoretically "correct" tool is useless if a developer can't easily implement it on any platform. While a future version will likely use more structured records (like HTTPS/SVCB), TXT provides a pragmatic, zero-friction starting point for everyone, today.

What This Means

• For Builders: You can now give your agent a permanent, discoverable address. The spec is final, the libraries are ready, and the documentation is live.
• For the Community: This marks the shift from an idea to a piece of usable public infrastructure.

The most valuable thing you can do now is build with it. Give your agent an address. Write a client that discovers one.

The official launch announcement is on X/Twitter, and sharing it is the best way to help grow the ecosystem:
[Link to your main Twitter thread]

All documentation, the rationale, and the live workbench can be found on the official site:
aid.agentcommunity.org

Link to AID

Thank you to everyone who has followed along so far. Let's get building.