Cursor has not announced anything yet, but the evidence for its next move is sitting inside its own web app. TestingCatalog found interface strings referencing a platform called “Cursor Review” that unlocks two new tabs, Codebase and Review, once access is switched on. The outlet frames this as a signal, not a confirmed release, and that distinction matters: nothing here comes from a Cursor blog post or press statement.
The stakes are still real even in sleuthing form. Origin is the code review layer Cursor previewed at its Compile conference in June, and if the tabs are close to shipping, Cursor is about to compete directly for a market that GitHub, Graphite, and CodeRabbit already sell into: reviewing code rather than writing it.
The two tabs split cleanly. Codebase handles syncing and managing repositories pulled from GitHub. Review is the piece TestingCatalog flags as more significant: a pull request pipeline that flags a developer for input only at the moments their judgment is required, so a person and a coding agent can clear a backlog of open PRs together. TestingCatalog’s signals put a possible launch inside days, well before the fall window Cursor originally pointed to when it first named Origin back in June.
Origin’s underlying case, per TestingCatalog’s reporting, is that GitHub’s review model assumes one reviewer working through one diff at a time in sequence. At Compile, Cursor put a number on the alternative: 22.6 commits landing in a single repository every second, a pace aimed at teams running fleets of background coding agents rather than individual engineers. That figure is Cursor’s own; independent throughput numbers for Origin have not surfaced.
The bottleneck argument is the interesting part. Generating code has gotten cheap enough that review, not writing, is where agent-driven engineering teams now get stuck. A pull request pipeline that pings a human only at decision points is a direct answer to that specific failure mode, and it is why GitHub’s own Copilot review features, Graphite’s stacked-diff workflow, and CodeRabbit’s automated review bots have all converged on the same problem from different directions. Cursor building this natively, rather than bolting it onto an existing GitHub workflow, is the differentiation it is betting on.
The tab structure also reveals a sequencing choice: land the review layer while repositories stay put on GitHub, then decide later whether hosting migration is worth pursuing. Cursor gains adoption from customers who have no intention of ripping their source control off GitHub in the near term, which is presumably why Codebase ships as a sync tool rather than a replacement.
This sits inside a larger roll-up. SpaceXAI recently pushed Grok Bot into beta, an app for desktop and mobile that hands agents a cloud machine of their own, letting them log into tools and finish assigned work without a human present, and TestingCatalog notes the product already carries references to Origin internally. SpaceX’s reported $60 billion purchase of Anysphere, the company that owns Cursor, is on track to close within the current quarter. Separately, a model tagged “Cursor Grok 4.6” briefly showed up in Cursor’s model picker before it was pulled. Two roadmaps that started apart are now visibly converging.
None of this is confirmed by Cursor, so treat the rollout date as directional rather than fixed. Engineering leads running background-agent fleets should watch for Cursor Review specifically, since a working PR triage layer would change where their teams spend review time long before any GitHub migration question comes up.
TestingCatalog reported these findings on August 11, 2026.
