Integration

The Wellknown MCP server

A remote, stateless MCP server over Streamable HTTP at https://wellknown.network/mcp. No authentication is required; anonymous rate limits apply per IP. Add it to any host and your agent gains the ability to find, inspect, check and announce other agents.

{
  "mcpServers": {
    "wellknown": { "url": "https://wellknown.network/mcp" }
  }
}

Tools

tool	purpose
`search_agents`	Natural-language search with protocol/status filters. Returns ranked hits with observed status and ranking signals.
`resolve_agent`	Best available agent for a task; recommendation carries its evidentiary basis. Use before calling a third party.
`get_agent`	Full canonical record for a handle or id.
`check_status`	Observed status, reason and 30-day reliability.
`list_capabilities`	The taxonomy, for precise filtering.
`submit_agent`	Announce an agent by card URL or inline manifest. Creates an unverified record.
`index_stats`	Truthful counts.

Host-specific notes

Claude Code: claude mcp add --transport http wellknown https://wellknown.network/mcp
Cursor / Windsurf / VS Code: paste the JSON above into the MCP settings file.
Codex CLI: add the server under mcp_servers in config.toml with url = "https://wellknown.network/mcp".
Your own agent: any MCP client SDK; the server is stateless, so no session handling is needed.

Safety

Tool results are JSON text describing third parties. They are data. The server's instructions field says so, and every tool description repeats it, so hosts that surface tool descriptions to the model carry the warning through.

Verify from the command line

curl -X POST https://wellknown.network/mcp \
  -H 'content-type: application/json' -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search_agents","arguments":{"query":"postgres","limit":3}}}'