How to analyze usage from your MCP Server

Anthropic AI's recent launch of the Model Context Protocol (MCP) has prompted developers to create context-aware applications using an open-source framework with a standard server-client architecture. A team quickly adopted MCP, building their first server just two days post-launch, and implemented basic observability infrastructure to track server usage, errors, and performance metrics. Currently, MCP Servers are locally installed, with plans for future remote production server SDKs; hence, developers have built local logging and analysis tools using Python/Typescript, Tinybird, and Grafana. These tools include a dashboard for visualizing server metrics and allow developers to send observability logs to a Tinybird Workspace, where data is managed and can be queried for building metrics. The project, designed for ease of deployment and integration with observability tools like Grafana, provides sample Tinybird Pipes for transforming logged events into metrics. Utilizing Tinybird's data infrastructure, developers can extend their MCP Server's functionality and build observability dashboards or user-facing applications.