Model Context Protocol · Streamable HTTP

One MCP endpoint. 20+ live data nodes.

Drop AIAgentNode into Claude Desktop, Cursor, or any MCP-compatible agent with one config line. Tools are tier-gated by your existing API key — no per-app keys, no rebuilds.

Tools exposed

Two tools. Every node.

list_nodes()

Discovery. Returns every available node's id, title, description, primary buyer, REST endpoint, and starting price.

No parameters.
search_records()

The primary query tool. Returns the canonical { source, source_id, type, discovered_at, payload } envelope from any node you're subscribed to.

node_id (required), limit, since (ISO-8601), query
Install

Three minutes to live data in your agent

Mint an API key in the dashboard, paste the config below into your MCP client, restart it.

Claude Desktop
~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "aiagentnode": {
      "type": "http",
      "url": "https://www.aiagentnode.io/api/mcp",
      "headers": {
        "Authorization": "Bearer aan_live_YOUR_KEY"
      }
    }
  }
}
Cursor
~/.cursor/mcp.json
{
  "mcpServers": {
    "aiagentnode": {
      "url": "https://www.aiagentnode.io/api/mcp",
      "headers": {
        "Authorization": "Bearer aan_live_YOUR_KEY"
      }
    }
  }
}
Any MCP client
Streamable HTTP
# Streamable HTTP MCP endpoint
URL:     https://www.aiagentnode.io/api/mcp
Auth:    Authorization: Bearer aan_live_YOUR_KEY
Method:  POST (GET/DELETE return 405)

# Available tools
list_nodes()                                 → catalog of all data nodes
search_records(node_id, limit?, since?, query?) → latest records, tier-gated
Same key, every client

Re-use your existing AIAgentNode key. No per-app credentials.

Tier-gated server-side

History windows, page sizes, and monthly quotas enforced inside the MCP server.

Same envelope as REST

Tool responses match /v1/* exactly — agents learn one schema.

Usage in your dashboard

Every tool call appears in your existing usage tracker.