Relvy, a Y Combinator Winter 2024 company, just launched to tackle a problem every on-call engineer knows: runbooks that sit in Notion gathering dust. Founders Kevin O'Toole and Eric Jiang, both former engineers at Samsara, built the platform to automate the boundary between what software can fix on its own and what actually needs human judgment. The idea came from their own time running large-scale infrastructure at Samsara, where they saw how static documentation rots while the systems it describes keep changing.
The Hacker News crowd confirmed the problem is real. Multiple commenters pointed out that maintaining dynamic runbooks is genuinely hard, and most teams default to wikis nobody updates. But tougher questions came around differentiation. Several people asked why you couldn't just use Cursor's agents with Model Context Protocols to spin up agents in a private cloud, giving them direct database and host access. Fair question. General-purpose cloud agents are getting better fast, and the line between specialized ops tools and general AI has started to blur.
Relvy's bet is that there's real value in a tool built specifically for incident response rather than relying on a general agent to figure out your operational stack. Whether that specialization wins over engineers already eyeing broader AI tools for the same job, such as botctl, is an open question.