AI
Cursor Just Pointed Its Bots at the Last Mile, and Shipping Code Grew a Supervisor
The most interesting AI coding launch this week had nothing to do with writing code. On September 23, Cursor published a blog post introducing two automated bots for Teams and Enterprise customers, tools the company describes as covering the last mile of shipping code. The writing assistant era is maturing into something more ambitious. Software that watches what happens after you merge.
The first bot, Rollouts, reads a pull request the moment it opens, maps out which systems the change touches, and writes a monitoring plan directly into the PR as a comment. The plan names the risks it spotted, the intended effect of the change, the signals it will check after deployment, and any instrumentation gaps that would make the change hard to verify. Engineers can edit the plan before the merge, and Rollouts works from the revised version.
After the deploy, Rollouts keeps watching. It connects to source control, the team's continuous delivery system, and telemetry providers including Datadog, Grafana, and Honeycomb, then compares live metrics against the predeploy baseline, tracking each environment separately. Every change gets one of three verdicts. Verified healthy, regression detected, or inconclusive. When it finds a real regression, it names the suspected change and notifies the author. Depending on configuration, it can open a revert pull request for review or hand the finding to a cloud agent for a fix. Humans still approve every merge and every rollback; the bot builds the case, the engineer signs it.
The second bot, Security Review, runs on every pull request and reads each change in the context of the full codebase, hunting the vulnerability categories that slip past ordinary peer review. Injection flaws, broken authentication on routes, credentials committed to the repository, unsafe deserialization, unvalidated redirects, and known issues in dependency changes. Each finding ships with a severity rating, an explanation of the attack path, and a one click fix. The updated bot averages 3.8 minutes per review, down from 4.8, a 21 percent improvement, and developers now accept its suggestions 60 to 70 percent of the time, up from 45 to 50 percent. Style and code quality reviews stay with the existing Bugbot.
The numbers that matter are the acceptance rates. Scan time is a vanity metric; engineers acting on findings is the metric that actually reduces shipped vulnerabilities, and the jump from roughly half of suggestions to roughly two thirds is the signal that the bot earned trust. Both bots activate from the automations tab in the Cursor dashboard and run on the team's existing telemetry stack, so teams skip buying separate tooling. Cursor is also offering trial credits for the next ten days, roughly 50 changes for Teams plans and 500 for Enterprise, a window that closes around October 3.
For builders, the takeaway is that AI assistance is climbing the stack. Autocomplete helped you type. Agents helped you write. Rollouts and Security Review aim at the anxious stretch after the merge button, the part of software work that has always resisted automation because it demands judgment across the whole system. The reviewer that reads your entire codebase in under four minutes is here. Your next release gets a supervisor.
Quick answers
What is this story about?
The most interesting AI coding launch this week had nothing to do with writing code. On September 23, Cursor published a blog post introducing two automated bots for Teams and Enterprise customers, tools the company describes as covering the last mile of shipping code. The writing assistant era is maturing into something more ambitious. Software that watches what happens after you merge.
Why does this story matter?
For builders, the takeaway is that AI assistance is climbing the stack. Autocomplete helped you type. Agents helped you write. Rollouts and Security Review aim at the anxious stretch after the merge button, the part of software work that has always resisted automation because it demands judgment across the whole system. The reviewer that reads your entire codebase in under four minutes is here. Your next release gets a supervisor.
Sources
- Pondero: Cursor ships Rollouts bot to watch deployments per PR
- Martin Cid Magazine: Cursor AI bots catch security flaws and halt bad deploys
- TerminalBlog: Cursor Just Became an Agent Fleet Manager
New to crypto? Read the crypto glossary, browse frequent questions, read our story, or explore the story archive.