Lightrun

Lightrun MCP connects your AI assistant to real runtime data from applications already running in production, staging, or development. Built on the Model Context Protocol, it lets Claude investigate live code behavior on your behalf—without changing code, rebuilding, or redeploying. **What you can do** - Discover connected agents, agent pools, and tags in your environment - Capture expression values and call stacks at specific code locations (snapshots) - Count how often a line executes - Measure execution duration between two lines (TicToc) - Sample custom numeric metrics from live code - Drill into raw measurement samples for deeper analysis **Why use it** Lightrun brings production-safe, on-demand observability into natural-language workflows. Instead of configuring probes manually, describe what you need and let the AI discover targets, place temporary instrumentation, and analyze results. Runtime data is subject to your organization’s PII redaction and permission rules. **How it works** Your AI assistant acts as an MCP client and connects to your Lightrun server at `/mcp`. The Lightrun MCP server translates requests into safe runtime inspections executed by Lightrun agents attached to your applications. **Authentication** - OAuth for interactive use (sign in with your Lightrun user) - API key (Bearer token with Dev scope) for agents and automation **Works great with Lightrun AI Skills** Pair this connector with reusable investigation workflows from the Lightrun AI repository—preflight checks, evidence collection, diagnosis, and handoff output for live runtime debugging. **Requirements** - A Lightrun account and server (SaaS: https://app.lightrun.com) - MCP access enabled for your organization - Lightrun agents connected to the applications you want to inspect