Relvy AI (YC F24) introduces an on-call runbook automation agent designed to reduce the burden on software engineering teams. The AI agent analyzes telemetry data and code at scale to diagnose production issues faster. Relvy tackles the challenge of autonomous root cause analysis, which benchmarks show struggles with accuracy, particularly on datasets like OpenRCA. Teams often use AI tools like Cursor or Claude Code with Datadog, but Relvy focuses on making investigations deterministic and context-aware.

Its tools detect anomalies in time series data, search logs, and reason about service dependencies without overwhelming the agent's context. AWS CLI commands can be automated for mitigation actions, requiring human approval. The system anchors investigations around runbooks, mirroring steps experienced engineers take, leading to quicker resolutions and less cognitive load for reviewers. Relvy is installed via Docker, Helm, or cloud signup, connects observability tools and code, and presents findings as interactive notebooks in its web UI.