Launching today

PostHog MCP Analytics
See how AI agents actually use your MCP server
1 follower
See how AI agents actually use your MCP server
1 follower
MCP Analytics shows how AI agents use your MCP server. Every tool call is logged with inputs, responses, duration and errors, plus error rate and p95 latency per tool. It also captures agent intent, the models agents report, and requests for tools you don't have yet. Failed calls land in Error Tracking, and every event works in PostHog dashboards, alerts and SQL. Open beta, free to start, no card needed.



Free Options
Launch Team

Render Sandboxes: a computer for every agentIsolated compute, persistent state and a real filesystem
Promoted
Hey Product Hunt! If you ship an MCP server, you probably have no idea how agents actually use it. Which tools get called, which ones fail, what the agent was trying to do when it gave up.
PostHog's MCP Analytics answers that. Add the SDK to your server and every tool call becomes an event:
inputs, responses, duration and errors for each call
error rate and p95 latency per tool
the agent's intent and the model it reports
requests for tools you don't have yet, basically a roadmap written by agents
Failed calls go straight into Error Tracking, and since it's all regular PostHog events, you can build dashboards, alerts or SQL on top. You can even query it over PostHog's own MCP from Claude Code or Cursor.
It's an open beta: the TypeScript SDK is pre-1.0 and Python is in beta. Free to start, no card.
MCP builders: what's the first thing you'd want to know about how agents use your server?