Skip to main content
Pulse is a drop-in telemetry SDK that captures tool-level usage and performance for your MCP server, then turns it into a real-time health signal in your MCPfy dashboard.
  • 📊 Tool-level telemetry
  • 🩺 Server health score
  • 🧠 Tool intelligence
  • 🎯 Result quality scoring
  • 🔌 Works with any MCP server - mcpfy-sdk, the official MCP SDK (@modelcontextprotocol/sdk for TypeScript, mcp for Python), FastMCP, or a server you didn’t write at all
MCPfy dashboard Overview section showing server health score, total calls, error rate, and health score and call volume trend charts

​
How it works

Every MCP transport - stdio, SSE, StreamableHTTP - funnels all JSON-RPC traffic through one seam: a send/onmessage pair in TypeScript, a read/write stream pair in Python. Pulse wraps that seam, so it sees every request and response regardless of which SDK built the server. For servers you don’t have the source for, mcpfy-proxy does the same thing from the outside: it becomes the process your MCP client spawns, runs the real command as its own child, and forwards every byte between them unchanged while classifying JSON-RPC messages on the side.
None of the three install modes below edit your files for you. Pick the one that matches your situation.

​
What can be tracked

Pulse classifies every JSON-RPC message crossing the wrapped seam and turns it into structured events - never raw content. This is what directly powers the dashboard’s Connections & Clients, Tools & Catalog Quality, and Server Signals sections.
Pulse never sends argument values, result content, resource contents, description text, or log message text - only their presence, size, and shape. The one exception noted above is tools/list, where description/param length is captured (never the text itself) to power description-quality scoring.

​
Get your API key

Every install mode needs an API key. Create one from your MCPfy dashboard (Dashboard → Telemetry), then set it as the MCPFY_API_KEY environment variable.

Open Telemetry Dashboard

Create an API key and see what the dashboard shows once events start flowing.

​
Which install mode do I need?

I'm building with mcpfy-sdk

Pulse ships bundled with mcpfy-sdk. Set MCPFY_API_KEY and restart your server - no code to write.

I wrote my own server and can edit its code

Install mcpfy-pulse and add two lines to your code, right before you connect your transport or run your server. See the TypeScript or Python guide.

I'm running someone else's server locally

A server from GitHub or npm, running under Claude Desktop or Cursor. Route its command through mcpfy-proxy in your MCP client’s config - no code needed. Covered in both the TypeScript and Python guides.
If apiKey (or MCPFY_API_KEY) is unset, every install mode is a complete no-op - the original transport or streams are returned untouched. It’s safe to leave the instrumentation in place across all your environments, including local dev.

​
Next steps

TypeScript

Install and configure mcpfy-pulse for Node.js MCP servers.

Python

Install and configure mcpfy-pulse for FastMCP or the official mcp SDK.

View telemetry in the dashboard

See tool-level usage, health scores, and quality signals once data starts flowing.