Mcpcat Typescript Sdk

MCPcat is an analytics platform built specifically for MCP (Model Context Protocol) server owners. It provides rich user analytics, session replay, issue tracking, and performance monitoring to give you visibility into how agents and users interact with your MCP server. By instrumenting your server with a lightweight SDK (available in TypeScript and Python), you can track every tool call, understand agent intent, and identify where agents get stuck. The platform is trusted by fast-growing AI startups and offers features like agent goals, funnel analytics, and error grouping. What makes MCPcat different is its focus on MCP-specific telemetry—it enriches sessions with agent intent and goal classification (optionally powered by LLMs), helping you understand not just what happened but why. It's built for developers who want to move from flying blind to data-driven roadmap decisions, with a free tier for experimentation and paid plans for scaling teams. Key features include session replay with step-by-step tool call inspection, agent goal enrichment for every session, error grouping with prioritization, per-tool performance monitoring, and custom user attribute filtering. The LLM Goals add-on provides intelligent session classification against custom goals, enabling funnel and drop-off analytics. The Growth plan ($160/month) includes 2,000 sessions per month and adds unlimited projects, team collaboration, and issue tracking. Enterprise plan offers custom volumes, exports to S3/Snowflake/BigQuery, SSO, audit logs, and priority support. Compared to general-purpose observability tools like Datadog or Sentry, MCPcat is purpose-built for MCP — it understands agent goals and intents natively, providing insights you can't get from generic APM. For MCP server builders, the free tier is a fast way to start debugging and understanding user behavior; paid tiers are reasonably priced for the specialized value.