Code review is the step where somebody reads a change before it ships. A developer proposes an edit, opens what the trade calls a pull request, and a colleague checks it for bugs and security holes.

It is slow and unglamorous work. Until recently it was also manageable, because roughly one human wrote each change and roughly one human read it.

That ratio has broken. Coding agents now work for hours at a stretch and open pull requests with barely any human involvement.

On Wednesday CodeRabbit said it had raised $143m to sit in the gap. Bloomberg reported the round before the announcement. Atomico and Smash Capital co-led it.

Chief executive Harjot Gill gave the post-money figure to Axios Pro as $1.5bn. He and Gur Singh founded the company in San Francisco in 2023. Gill had sold an earlier startup, Netsil, to Nutanix in 2018.

Nobody is reading what the agents write

The company sources the problem from outside itself. GitHub is on pace to record 14 times more commits this year, on its chief operating officer’s own figures. Among firms in the top decile of coding-agent adoption, agents open 35% of pull requests, according to Jellyfish.

Demand for the agents keeps outrunning supply. GitHub Copilot paused new signups rather than let usage run away from it.

“AI made code plentiful,” CodeRabbit wrote on X. “Trust is now the bottleneck.”

The product reviews changes automatically. It flags bugs, security flaws and maintainability risks, and it explains what each change does. It runs more than two million reviews a week for over 17,000 customers, the company said in its funding release. Nvidia, BMW, Adyen, Indeed, JFrog and Trivago are among them.

A European fund co-led an American round

Atomico is a London firm, and it does not often co-lead US growth rounds of this size. Partner Luca Eisenstecken takes a board seat.

“As AI becomes critical infrastructure for the global economy, organisations will increasingly need independent governance layers that can validate software regardless of which model produced it,” Eisenstecken said.

The money follows the customers. CodeRabbit has opened an office in Moorgate in London and now employs 50 people across the UK and the EU. It plans to double that European headcount, then move into Japan and Singapore.

BMW i Ventures also joined the round. Managing partner Kasper Sage said carmakers depend on reliable software and efficient engineering teams. BMW is already a CodeRabbit customer, so the fund is backing a product its own engineers use.

Several backers sit on both sides of the ledger. Datadog is a new investor. Nvidia was already one, and it is a customer as well.

That Nvidia tie is technical, not only financial. CodeRabbit post-trained Nvidia’s Nemotron 3.5 Lightning model to handle one of its highest-volume routing jobs. Existing backers CRV, Scale Venture Partners, Flex Capital and Pelion Venture Partners all returned.

The bigger claim is that Jira is finished

CodeRabbit announced a product category alongside the money, and called it Agentic Change Management. The founders’ blog post puts it more bluntly. “Issue tracking is dead.”

The argument runs like this. Trackers such as Jira were built when writing code was expensive, so teams planned first and built second. AI reverses that order. Code now exists before anyone has decided whether it is worth having.

Anyone can start a change, too. OpenAI has pushed Codex at non-developers for much the same reason.

Three new tools follow from that. Triage scores incoming pull requests by value, urgency and risk, then routes the low-risk ones into automation. Change Stack groups a large agent-written diff into layers a human can follow. CodeRabbit Security scans code that has already shipped.

Security is what the cheque was really for

Large language models cut the cost of finding vulnerabilities. That helps defenders and attackers equally. It is why the security product arrived with the round rather than after it.

“Security now requires a continuous system that can reason across the repository, verify whether a threat is real, and move the fix back through the development workflow,” said Brad Twohig, founding partner at Smash Capital.

The desk has covered the failure mode before. Lovable’s vibe-coding platform ran into a security crisis of its own.

The supply chain shows the same shape. An npm worm spread through developer packages this year.

The market is smaller than the company

Here is the awkward number. The global code review services market is worth about $3.04bn this year and should reach $4.79bn by 2032, on 360iResearch figures cited by Tech Funding News.

CodeRabbit is valued at $1.5bn. That is roughly half the entire market it currently sells into.

The valuation only works if code review stops being a developer tool and becomes infrastructure. Gill says the pull is there. “It’s almost like a land grab moment for us,” he told Bloomberg.

Rivals are grabbing too. Qodo raised a $70m Series B led by Qumra Capital, Greptile raised a Series A at a $180m valuation, and CodeAnt AI closed a $2m seed.

Cost is the other pressure. Gartner expects AI coding to cost more than developers by 2028, and a review layer adds to that bill rather than trimming it.

What the company did not disclose

Revenue grew more than five times year over year, CodeRabbit said. It did not say what it grew from. The multiple therefore carries no absolute figure.

Total funding is about $231m by Tech Funding News arithmetic, counting a $16m Series A in 2024 and a $60m Series B last September. Bloomberg puts the total at more than $200m.

The company will also spend $10m over 12 months keeping the tool free for open-source maintainers, Reuters reported. That covers 150,000 projects.

What would settle it

Two things are checkable rather than rhetorical. The first is the London headcount, which the company says will double from 50.

The second is the real claim. “Issue tracking is dead” is a big sentence, and it stops being marketing on the day a named customer says it dropped Jira for CodeRabbit Triage.

Until then this is a very well funded company selling a fix for a problem the AI industry created, to the same industry that created it.

Get the TNW newsletter

Get the most important tech news in your inbox each week.