↓ Documentation

How to use A2A Search

What is A2A?

The Agent2Agent protocol is an open standard (Google + 50+ partners) for AI agents to discover each other, delegate tasks, and exchange data through a uniform HTTP+JSON interface.

A2A Search is a public catalog of A2A-compatible agents. Anyone can publish. Anyone — humans or other agents — can search. We host nothing about the agents themselves: just metadata pointing at their public Agent Cards.

Publishing your agent

Your agent must serve a valid Agent Card at /.well-known/agent.json. Minimum structure:

{
  "name": "My Agent",
  "description": "What it does",
  "skills": [
    {
      "id": "schedule_meeting",
      "name": "Schedule meeting",
      "description": "Book a meeting with the user",
      "inputSchema": { "type": "object", ... }
    }
  ]
}

When ready, visit Submit and paste the URL. We'll fetch, validate, and publish in one step.

Discovering agents from your bot

The fastest way: connect your bot to our MCP endpoint.

MCP endpoint
https://a2a-search.com/mcp

From any MCP-compatible client (Claude Desktop, Anthropic API tool-use, custom agents) you immediately get three tools:

  • search_agents — free-text search + optional category
  • get_agent — full Agent Card by id
  • list_categories — with counts

REST API

If you prefer plain HTTP:

GET https://a2a-search.com/api/agents?q=research
GET https://a2a-search.com/api/agents/{id}
GET https://a2a-search.com/api/categories
POST https://a2a-search.com/api/agents
     { "url": "https://my-agent.com", "categories": ["x"] }

Health checking

We re-fetch every registered agent every 6 hours. After three consecutive failures, status changes to offline. A successful check returns it to live. If you provide a contact email at submission, we'll log when your agent goes offline (notifications are not yet wired but logs are kept).