Hey Product Hunt! 👋 We've talked to dozens of companies building MCP servers, and not a single one could answer: "What are users actually doing with your MCP servers?" Companies are spending millions on MCP servers but they're flying completely blind. They can't answer basic questions like: - When did AI agents hit errors? - What patterns cause failures? - How are agents actually using the tools in practice? Normally, you'd use Datadog or Sentry for this. But MCP is so new that there wasn't an easy way to connect these tools. So we built one. MCPcat gives you two things: 1) **Free, open-source telemetry** that connects your MCP server to the observability stack you already use (OpenTelemetry, Datadog, Sentry). Your data never touches our servers. 2) **AI agent session replay** (think LogRocket for MCP)—see exactly how AI agents interact with your tools, understand their intent, and debug issues in context. This part is optional but game-changing for product teams. Everything is open source because we believe developer tools handling potentially sensitive data should be transparent. We're actively building based on community feedback. If you're building an MCP server, we'd love to hear what observability challenges you're facing! - Kashish & Naseem

MCPcat

User analytics and telemetry for MCP servers

User analytics and telemetry for MCP servers