Cursor acquired Firetiger. A month later, it launched a bot that tracks code changes from PR to production.
The New Stack Paul Sawers
Cursor launched Rollouts, a bot that watches code after it ships. It’s trying to catch production bugs and bad deploys before users do.
Based on reporting by The New Stack, Paul Sawers — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Cursor is pushing its AI tools past the point where most coding demos stop. The company has introduced Rollouts, a new agent that follows a change from pull request to production and checks whether it behaves the way it should. That matters because writing code has gotten easier; proving that code is safe after it ships is still the messy part.
The launch comes just over a month after SpaceX closed its $60 billion acquisition of Cursor. Right before that, Cursor bought the Firetiger team, a three-year-old startup focused on monitoring software changes from pull request through deployment. Firetiger co-founder and CEO Rustam Lalkaka had been making the same basic argument before the deal: agentic coding has made changes cheap to create, but the cost and risk of deploying them has barely moved.
Rollouts is the first visible result of that acquisition. Cursor says the bot builds a monitoring plan as soon as a pull request opens, working out what systems could be affected, what should be watched, and where the gaps in instrumentation are. Developers can review and edit that plan before anything reaches production. After deployment, Rollouts compares logs, metrics and traces against the plan, with separate checks for staging and production. Each deployment ends up marked as verified healthy, regression detected or inconclusive.
And if something looks wrong, Rollouts can point to the suspected change, notify the developer, and either open a revert pull request or hand the issue to a Cursor cloud agent for a possible fix. It still stops short of making the final call on its own. The human is still in the loop where it counts, which is probably a good thing when the bot is deciding whether checkout is broken in one region or everywhere.
Cursor is also upgrading Security Reviewer, a bot that scans pull requests for security problems, broken authentication, privacy issues, tool auto-approvals and prompt-injection attacks. The company says review time has dropped from 4.8 minutes to 3.8, and acceptance of its comments has climbed from roughly 45–50% to 60–70%. Both bots are available in the Automations tab for Teams and Enterprise customers.
My take — AI-written commentary, not fact-checked reporting
This is the right bet, and also a neat admission that most AI coding hype has been obsessed with the easy half of the job. Open models can write code all day; the real moat is deciding what that code did once it hit production. Cursor is building the boring, defensible part, which is usually where the money is.
Read more about this at: The New Stack