CodeRabbit Raises $143M Series C to Police the Code Agents Are Writing
CodeRabbit raised $143 million in a Series C at a $1.5 billion valuation, co-led by Atomico and Smash Capital. New money from BMW i Ventures, Datadog, Hirtle Callaghan, Scenic Management and SineWave Ventures, with CRV, Flex Capital, Scale Venture Partners, Harmony Partners, Pelion and Engineering Capital returning. Atomico partner Luca Eisenstecken joins the board. It comes less than a year after a $60 million Series B, and revenue grew more than 5x year over year.
The business is AI code review, and the volume numbers are the argument: more than two million code reviews a week across 17,000-plus customers including Adyen, Indeed, BMW, Nvidia, JFrog, Trivago and Campfire. Alongside the round they launched Agentic Change Management.
The thesis is almost embarrassingly simple. Agents write vastly more code than humans ever did, and every line still needs review, but the review capacity didn't grow. Somebody was going to get paid to sit in that gap. What's interesting is who: not a model lab, not an IDE, a review company. Generation got commoditized in about eighteen months. Judgment about generated code did not.
Worth noting how this rhymes with Zed Delta from yesterday, which is attacking the same problem from the other end by rebuilding the review surface itself instead of automating the existing one. Two well-funded bets that the pull request, designed for humans reviewing humans at human volume, is the actual bottleneck in agent-written software. The BMW i Ventures and Datadog checks tell you where the demand is coming from, and it isn't startups.
Details at coderabbit.ai.
← Back to all articles
The business is AI code review, and the volume numbers are the argument: more than two million code reviews a week across 17,000-plus customers including Adyen, Indeed, BMW, Nvidia, JFrog, Trivago and Campfire. Alongside the round they launched Agentic Change Management.
The thesis is almost embarrassingly simple. Agents write vastly more code than humans ever did, and every line still needs review, but the review capacity didn't grow. Somebody was going to get paid to sit in that gap. What's interesting is who: not a model lab, not an IDE, a review company. Generation got commoditized in about eighteen months. Judgment about generated code did not.
Worth noting how this rhymes with Zed Delta from yesterday, which is attacking the same problem from the other end by rebuilding the review surface itself instead of automating the existing one. Two well-funded bets that the pull request, designed for humans reviewing humans at human volume, is the actual bottleneck in agent-written software. The BMW i Ventures and Datadog checks tell you where the demand is coming from, and it isn't startups.
Details at coderabbit.ai.
Comments