Spanly

Spanly is an observability platform built specifically for Model Context Protocol (MCP) servers. It captures every tool call, resource request, and prompt exchange — with full JSON-RPC payloads — across any language and transport (stdio or HTTP). The service works at the protocol level, wrapping your server via a CLI or SDK with less than 1ms overhead per request. No code changes required.

The platform provides real-time traces, error tracking, and performance dashboards that show latency percentiles, error rates, and top slowest operations. It complements existing APM tools like Datadog or Sentry by focusing on MCP-specific data: you see the full request and response packets, client and server versions, and can correlate traces with APM spans. Spanly also offers multi-tenancy through headers or JWT claims, letting you separate telemetry by client or tenant. Data residency is built in — pick US or EU when you create a project, and all telemetry stays there.

Engineering teams shipping MCP servers in production will get the most value. The on-call engineer can see exactly which tool call failed and why, without patching together spans from different systems. It also helps catch bad deploys by tagging every request with server and client version. If you're running MCP servers for AI coding assistants like Claude, Cursor, or Copilot, Spanly gives you the visibility to debug issues and track performance across deployments.