Timing middleware is a lightweight component in the request-processing pipeline that measures how long each request takes to execute. When runtime.middleware.timing.enabled=true, the MCP Server activates a lightweight middleware that records precise duration metrics for each request. These metrics help administrators and developers analyze latency and identify performance bottlenecks. Timing middleware introduces minimal overhead and is recommended for production environments.

Best practice: Combine timing middleware with structured logging to compute latency percentiles such as P50 and P95 externally using log aggregation tools. Performance percentiles are statistical measures that indicate how a system performs for a given percentage of requests. They help you understand not just the average performance but also the distribution of response times across all requests. Average response time can hide outliers. Percentiles provide a clearer picture of user experience by showing how most requests behave and identifying slow responses.