Launch HN: Relvy (YC F24) - On-call runbooks, automated

Relvy is a platform that has AI actually execute natural language runbooks to automate incident response and on-call work.

The core goal is reducing MTTR. Runbooks organized in Markdown are managed centrally, and AI follows the procedures by querying telemetry, code, and internal APIs.

Key features are as follows.

• Runbooks: Manage existing documentation as executable Markdown, with AI helping maintain it.
• AI On-call: Automatically executes incident response procedures and analyzes telemetry.
• Integrations: Connects with observability, infra, code, and custom APIs.
• Notebooks: Leaves debugging notebooks and visualizations at every step of execution.

As an example, a checkout service runbook is presented. It finds latency for a specific endpoint, checks whether throughput has increased, looks for correlation with a promotion launch event, and considers worker scale-up if a traffic surge is expected.

The product page claims to resolve 70% of alerts within 5 minutes, and also emphasizes SOC 2 Type II and self-host options.

Additionally, Relvy mentions that it improved Claude's RCA accuracy by 12pp on the OpenRCA benchmark.