Anysphere, maker of the Cursor AI coding assistant, announced today that it has acquired Graphite, a fast-growing startup known for its Paleolithic-looking summaries in AI-powered code review and its “stacked pull request” workflow. No terms were revealed, but Axios reported the deal price “blew out” Graphite’s previous valuation of about $290 million, which it saw earlier this year when it raised a $52 million Series B — indicative both of the heat in AI developer tools and Anysphere’s desire to create an end-to-end platform with closely knit links across those kinds of services.
Why Graphite Fits Cursor’s Roadmap for AI Code Reviews
AI-prompted code can be fast, but it isn’t always tidy. Several studies and analyses in the industry have found higher rates of bugs or security vulnerabilities in code written by large language models without careful review. It’s that gap — the one between speed and accuracy, reliability, etc. — that Graphite has focused on so successfully. Its use of the stacked pull request model allows developers to divide large changes into smaller PRs that depend on each other and can be reviewed, tested, and merged one at a time.
The methodology, which has been used inside organizations including Meta and Google with “stacked diffs,” reduces the friction of review and minimizes cycle times by keeping changes atomic. Rather than one giant PR that sits waiting for approval (or failing to be LGTM’d), teams can move a string of small, interconnected changes along CI and CR in parallel. For AI-assisted coding, where there are frequent iterations that touch many files, stacked PRs can materially improve developer velocity without spiking risk.
Cursor already has automatic review through its Bugbot, but the depth of workflow engine and developer-first UI in Graphite focused on sequencing, visualizing dependencies, and resolving conflicts hits higher heights. Fold that into Cursor’s code generation and refactoring wizardry and Anysphere inches itself closer toward a closed-loop system: propose code, reason about its impact, generate test cases for the code you’re going to write, shepherd those changes through to merge — all with AI in the loop.
Inside the Deal and the Shared Investor Network
The companies have some notable investor DNA in common. Graphite had been backed at the seed stage by Neo, Ali Partovi’s early-stage firm, and Accel and Andreessen Horowitz are backers of both Graphite and Anysphere, according to data from PitchBook. Graphite’s Merrill Lutsky, Greg Foster, and Thomas Reimers first met Michael Truell, co-founder and CEO of Anysphere, through the Neo Scholar program, which sources (among other types of talent) college engineers who are considering founding a company.
Anysphere has been on an acquiring tear. In recent months, it purchased Growth by Design, a strategy firm for tech recruiting companies, and earlier this year it took over talent from AI-powered CRM startup Koala in what was reportedly a deal with a $129 million post-money valuation, according to PitchBook. The company itself was valued at approximately $29 billion late last November, which speaks to the resources driving its roll-up strategy in AI software tooling.
A Crowded AI Code Review Market Is Taking Shape
Anysphere’s stack stationed itself squarely in competition amid the emergence of Graphite. Another AI-driven code review player, CodeRabbit, was valued at $550 million in September, and even old-school Greptile raised a $25 million Series A this fall. In the meantime, incumbents on these platforms are closing in on similar frontiers: GitHub is expanding Copilot into code review and test generation, and large cloud providers are embedding AI agents in IDEs that can suggest diffs and explain changes.
And the distinction is less and less a matter of model quality, and more and more about depth of workflow. Features like stacked PRs, automated test generation, intelligent reviewer assignments, and CI-aware suggestions are increasingly becoming table stakes for enterprise engineering leaders who watch DORA metrics such as deployment frequency and change failure rate. By combining Cursor’s coding agent with Graphite’s sequencing and review automation, Anysphere is betting that it can compress the time from prompt to production while keeping guardrails in place.
What Stacked Pull Requests Enable for Teams and AI
You can imagine a multi-file refactor motivated by an AI suggestion: rather than ship one giant PR, the developer splits the work into a chain (“first add this interface, then go migrate that module over to it,” and so on), each step is reviewed and tested in sequence. When a step fails CI, the stack stops there without standing in the way of unrelated improvements. For managers, it lowers the clustering of risks and gives better observability into where reviews get stuck and why.
And Graphite simplifies the duties that frequently go neglected in AI demos: merge conflict resolution, rebase automation, and policy checks. Combined with Cursor’s Bugbot’s code understanding, these two capabilities could lead to a policy-driven reviewer who suggests changes, runs tests, fixes linting, and progresses the PR stack automatically as things occur.
What Changes for Developers Using Cursor and Graphite
For the short term, expect deeper integrations on Cursor and Graphite: inline summaries for review selectively linking to generated diffs; prompts that know what stacks are, that will recommend your next PR (“next,” as in dare I say “reviewers”?) and more tailored analytics around review time and flake rates. And for security and platform teams, the combined offering could allow organization-wide rules — for example, that all tests be required or negotiated by a particular person or group, or that certain risk thresholds be adhered to — to be enforced automatically by the AI agent as it progresses a stack toward merge.
The larger question is the ecosystem impact. If Anysphere can make stacked workflows turnkey and model-agnostic across GitHub repositories, it puts pressure on rivals to rise past “AI comments” into full change management. That competition should help engineering teams that have been challenged by endemic review backlogs and growing pressure to ship faster without sacrificing reliability.
Bottom line: by acquiring Graphite, Anysphere isn’t just bolting on yet another AI helper — it’s weaving the workflow fabric that encases code generation. In a market where winners are judged by shipped results, not smart-sounding soundbites, that integration may be more important than any one model upgrade.
