KubeMQ
DeployScenariosAI-agent fabric

Connect an MCP host

Point Claude Desktop, Cursor, or any MCP host at KubeMQ's built-in MCP endpoint, list the tools, and drive messaging from your LLM.

KubeMQ must be running — see the Quickstart (steps 1–2 get you there; instant if it's already up).

KubeMQ's MCP gateway is already live on :9090 the moment the broker starts — there's no connector to enable and no flag to flip. Point any MCP host at the endpoint and it can discover and call KubeMQ's tools right away.

1 · Point your MCP host at KubeMQ

Add a server entry to your MCP host's config. For Claude Desktop this is claude_desktop_config.json (Claude Desktop → Settings → Developer → Edit Config); Cursor and other hosts use an equivalent MCP server list:

claude_desktop_config.json
{
  "mcpServers": {
    "kubemq": {
      "url": "http://localhost:9090/mcp"
    }
  }
}

The url must include the /mcp path — it's the only required field for an unauthenticated server. Restart the host so it picks up the new entry.

2 · Verify

Cross-check from both ends:

  • From the serverkmq mcp list prints every MCP tool KubeMQ has registered.
  • From the host — start a new conversation and ask it what tools it has; KubeMQ's tools (for example queue_send) should be in the list.

Then make one call as proof: ask the host to send a message to a queue channel — for example, "send hello to the orders queue." A successful call returns the tool's result text with no error.

Didn't work?

  • Wrong endpoint/URL — confirm the url is exactly http://localhost:9090/mcp (the /mcp path is required) and that port 9090 is reachable.
  • Host config — most MCP hosts need a full restart, not just a reload, to pick up a new or changed server entry; check the host's developer/logs panel for a connection error.
  • Auth mismatch — if the server has JWT auth enabled, add an Authorization: Bearer <jwt> header to the host's server entry; without it, calls fail even though the endpoint is reachable.

Go deeper

Was this page helpful?

On this page