Your AI Agent Can Now Ask Production About the Past | Lightrun

Most questions about production come after the fact. Something behaved strangely an hour ago, and now everything looks normal.

In this demo, Gidi Freud introduces runtime history, powered by the new Lightrun Runtime Collector. Your AI agent searches the snapshots Lightrun already captured in production, so when a question comes up later, the evidence is already there.

Watch an AI coding agent investigate one minute of production in two parts:

What did the service see? Roles, permission IDs, and users, pulled from 22 captured hits
Why did some requests fail? The one affected user, the upstream error the handler hid, and why every other request went through

The whole investigation runs on existing captures, with zero new snapshots. When runtime history doesn’t hold the answer yet, the agent places a live snapshot in the same workflow.

Runtime history is included in the Lightrun Runtime Sensor, starting with Lightrun 1.97.

Runtime history docs: https://docs.lightrun.com/mcp/runtime-history/
Connect your agent to Lightrun MCP: https://docs.lightrun.com/mcp/mcp-quickstart/