Developer

Agentmatica MCP server

Agentmatica is itself a Model Context Protocol server. Point Claude, Cursor, or your own agent at one URL to read the docs, compare plans, and inspect your agent.

Endpoint

JSON-RPC over HTTP POST — no SSE or long-lived connection needed.

POST https://agentmatica.com/api/mcp

Every request is answered with plain JSON, so it works on serverless hosts. A GET returns 405. Supported methods: initialize, ping, tools/list, and tools/call. Notifications get an empty 202.

Tools

All tools are public, read-only and need no header. An API token is accepted but there are no agent-specific tools yet; a bad token gets 401 / -32001.

• list_plans — every pricing tier with its price.
• compare_plans — left and right tier names side by side.
• search_docs — search this documentation by title and summary (query).
• get_doc_page — one docs page by slug.
• create_checkout_link — a full link for a plan: the visitor signs up or signs in and lands on that plan's checkout. Nothing is charged until they confirm.

Connect from a client

Claude Desktop / Claude Code style config:

{
  "mcpServers": {
    "agentmatica": {
      "url": "https://agentmatica.com/api/mcp",
      "headers": { "Authorization": "Bearer amk_live_…" }
    }
  }
}

Or call it directly:

curl https://agentmatica.com/api/mcp \
  -H "Content-Type: application/json" \
  -d '{ "jsonrpc": "2.0", "id": 1, "method": "tools/list" }'

Training endpoint

A second endpoint, https://agentmatica.com/api/mcp/train, lets a signed-in merchant train their agents from ChatGPT, Claude or any MCP client. It uses OAuth 2.1 sign-in (no API key) and offers list_agents, get_training_guide, list_rules, check_rules, save_rules, update_rule, set_rules_enabled and delete_rules. See Train from ChatGPT or Claude.