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

Relvy, a YC F24 startup, says it wants to shorten mean time to resolution by putting AI in charge of your on-call runbooks. Short story: feed it Markdown runbooks and telemetry and it will execute the steps. Sounds like having your most experienced engineer on-call 24/7. Too good to be true? Maybe — it has been reported that early customers are already using the product in production.

What it does

Relvy advertises an AI that executes natural‑language runbooks, analyzes telemetry and code, and can be extended with custom connectors to your internal APIs and tooling. Runbooks live centrally in executable Markdown; you can import existing docs and the company says the AI will help maintain them. It has been reported that Relvy provides collaborative debugging notebooks with rich visualizations for every execution, and that its foundational tools let it parse observability data at scale to make decisions during incidents.

Why it matters (and what to watch)

The pitch lands in a crowded, hungry market: SRE teams are stretched thin, and anything that reliably cuts MTTR gets attention. Relvy’s claim that it can reliably perform “real steps from production runbooks” — and thereby give engineers more time to build — is promising, but it raises familiar questions about trust, auditability, and safety. Forward Deployed Engineers from Relvy allegedly tune runbooks for customers, which might ease adoption, but teams will want clear logs, approval gates, and rollback paths before handing the keys to an AI. More vendors will chase the same promise: faster incidents, fewer midnight calls. Relvy’s site is relvy.ai for curious teams who want to poke the toolbox themselves.

Sources: relvy.ai, Hacker News