Get started with TrackMCP

Install the SDK, create a workspace key, wrap your server, and deploy. Data appears in your dashboard after the first telemetry flush.

# TypeScript
npm i @trackmcp/sdk
# Create a key at https://app.trackmcp.com/dashboard

          # Python
          python3 -m pip install trackmcp
# Create your key at https://app.trackmcp.com/dashboard

import { withTrackMCP } from "@trackmcp/sdk";
import { server } from "./mcp";

export default withTrackMCP(server, {
  apiKey: process.env.TRACKMCP_KEY,
  service: "acme-mcp-server",
  environment: "production",
});

That's the whole setup. The wrapper observes the MCP protocol layer and tool calls without changing your tool implementations. Errors returned inside a successful 200 OK are preserved in the captured result.

MCP clients first initialize a connection and discover the server catalog. They then call tools, receive results, and may retry or stop. TrackMCP observes this lifecycle at the server boundary and sends small batched telemetry events to your workspace.

Use the SDK that matches the language of the process hosting your MCP server. The wrapper is additive: it records telemetry without changing tool definitions or the responses your clients receive.

After starting the instrumented server, make one real tool call from your MCP client. Return to the dashboard and choose Check for data. A successful connection shows an event, a session, the discovered catalog, and the client name when the client provides one.