Hey guys, Mr. Technology here.
Cursor added two new bots to its Teams and Enterprise plans on Sep 23: Rollouts (a deploy-and-monitor agent that watches every change from PR to production, writes a monitoring plan, and flags regressions) and Security Reviewer (a PR-scoped security agent that reads code like a security engineer — injection, authn/authz bypasses, secrets in source, SSRF, unsafe deserialization, dependency vulnerabilities). Both are gated to Teams ($40/user/mo) and Enterprise today; both are powered by the team that built Firetiger, which Cursor acquired six weeks earlier. (Rollouts and Security Review blog post, changelog entry, Firetiger joins Cursor, Cursor pricing)
This is a documentation-comparison report. Every capability claim below is sourced verbatim from the Cursor blog post, the changelog entry, the pricing page, or the Firetiger acquisition post. I have not enabled either bot against a production repo; the operational mechanics below are what the documentation says, not what I tested.
Two bots, same launch window, same plan gating:
1. Rollouts attaches a monitor to every PR and watches the change as it deploys. Before merge, it reads the diff and writes a "monitoring plan" as a PR comment — risks it sees, the effect the change is meant to have, the signals it will check, gaps in instrumentation. After deploy, it compares the signals in the plan against the pre-deploy baseline per environment, names the change it suspects when it detects a regression, and depending on configuration pings the author, opens a revert PR for review, hands the finding to a cloud agent for a fix, or (per the changelog) pauses a progressive rollout. It does not merge or roll back on its own today. 2. Security Reviewer runs on every PR, reads the change in the context of the codebase, and posts one review comment per PR with severity, attack path, and a proposed fix. Draft PRs are skipped. Bugbot still owns style and quality; Security Review owns exploitable bugs.
Both are gated to Teams and Enterprise plans today. Per the changelog entry, the launch includes a 10-day window of usage credits: roughly 50 changes for Teams customers, 500 for Enterprise — both measured as monitored PRs (Rollouts) or reviewed PRs (Security Reviewer).
The changelog also commits to a roadmap: feature-flag integration so Rollouts can ramp and unramp traffic directly, awareness of release trains and deploy freezes, and team rules for Security Reviewer so customers can encode rules like "client external calls must go through X" or "table Y is never queried from a request handler."
Cursor now ships production-side agent capabilities, not just code-writing agents. Through 2025 and the first nine months of 2026, Cursor's agent surface was overwhelmingly on the pre-merge side — Composer, the Cloud Agents that run in Cursor's own environment, Bugbot for code review, Background Agents for long-running tasks. Self-Hosted Machines (Sep 2) extended the execution surface to inside the customer's network; Rollouts and Security Reviewer extend it past the merge. The agent now follows the change into production.
Rollouts is a structured, plan-first monitoring agent — not just "watch deploys." The PR-opens → writes monitoring plan → post-deploy verification loop is the substantive new shape. The monitoring plan is editable; Rollouts uses the customer's version. Per-environment verdicts ("verified healthy," "regression detected," "inconclusive") replace the binary green/red of a deploy hook. Rollouts tells intended effects apart from regressions, so a deliberate latency spike for a load test does not page anyone. It also flags missing instrumentation before merge — described as "the most common reason a bad change goes unnoticed."
Security Reviewer is a context-aware PR reviewer — not static analysis. The Cursor blog post is direct on the framing: "Static analysis pattern-matches: it flags every string concatenation near a SQL call and misses the authorization check that stopped running after a refactor. Security Review reads code the way a security engineer does: where does user input enter, where does it end up, what does it pass through on the way." Out-of-the-box coverage: SQL/command/template/LDAP injection, broken authn/authz on new and changed routes, secrets committed to source, SSRF and unvalidated redirects, unsafe deserialization, dependency changes that pull in known vulnerabilities, and insecure defaults in infrastructure/config.
This is the Firetiger acquisition shipping its first product. Firetiger built agents that operate production systems — monitoring rollouts, catching regressions, investigating incidents, passing findings back to coding agents. The Aug 13 "Firetiger joins Cursor" blog post by Maxime Prades (now-Cursor team) is direct: "You'll see Firetiger's work show up across Cursor as we build more of the path from writing code to running it in production. This is part of a broader investment in long-running, autonomous, context-aware agents for teams. That work also includes Cursor Origin, our Git forge built for the agentic era, and soon Change Monitors, which watch deployed changes and flag problems as they appear." The Sep 23 launch is the "Change Monitors" promised in August; the author byline is Rustam Lalkaka, Firetiger's co-founder.
Plan tiering matters — Individual and Hobby are out. The Rollouts and Security Reviewer changelog entry says it explicitly: "Rollouts and Security Reviewer are available today on Teams and Enterprise plans." The Cursor pricing page confirms the gate — Teams is $40/user/mo with Cloud Agents and automations plus centralized billing and SAML/OIDC SSO; Enterprise is custom and adds pooled usage, SCIM, repo/model/MCP access controls, audit logs, and the AI code tracking API. Individual ($20/mo Pro/Pro+/Ultra) and Hobby (free) do not get either bot. For an Individual customer, there is no work-around in this launch.
Integrations are real product surfaces, not roadmap items. Rollouts connects to Origin or GitHub for source control, your continuous-delivery system for deploy events, and Datadog (and "other telemetry providers") for signals. Security Reviewer connects to the source repositories you enable from the dashboard and respects dismissals — dismiss a finding with a reason and Security Reviewer won't raise it again on that PR. Team rules are enforced on every PR once configured.
For the engineering team that already runs Cursor: Rollouts closes the gap between "PR merged" and "verified healthy in production." If you have been spending on-call time investigating "which of eleven changes broke checkout," Rollouts is the structured answer. It is also a forcing function for instrumentation gaps — the PR-comment monitoring plan surfaces them before merge, which means your team's observability debt gets visible at code-review time instead of during incidents. Security Reviewer adds a parallel, exploitable-bugs-only review layer that does not duplicate Bugbot's style/quality role; if your team has been running Cursor without a separate security review pass, this is the cheapest path to one.
For the platform team that owns observability: expect to be the connector. Rollouts will need read access to your deploy events, your logs/metrics/traces, and (soon) your feature-flag system. Datadog and Grafana and Honeycomb are first-class; everything else is "other telemetry providers" that will likely need a thin adapter. Your paging rotations still own the "regression detected" verdict until a human reviews the Rollouts-authored revert PR.
For the security team that owns PR review: Security Reviewer is a complement, not a replacement, for whatever your team runs today (Semgrep, Snyk, internal review). It reads code contextually rather than pattern-matching, which is the gap every static analyzer admits. The trade-off: dismissals stick per-PR ("won't raise it again on that PR") — use the team-rules surface for cross-team policy instead of per-PR dismissals.
For the founder or VP Engineering weighing Cursor's trajectory: the product shape is no longer "a coding tool"; it is a CI/CD-adjacent agent platform with a coding-tool front door. Cursor is competing with Datadog for incident context, with Snyk and Semgrep for security review, and with GitHub Actions for the deploy hook — while still competing with Claude Code and Codex for the coding step. The Firetiger acquisition is the visible proof that this is deliberate.
For the open-source maintainer or hobby developer on Individual: nothing changes today. Both bots are Teams/Enterprise. If your team is on Individual and you want either capability, you are paying the $40/user/mo jump (or a custom Enterprise quote) for it.
Every capability claim above is sourced from one of the following primary sources:
The Cursor changelog hash signal sourcechange-cursor_changelog-f9c74f9c3be1 first triggered on 2026-09-26 05:09:54 UTC; the cursor_pricing signal sourcechange-cursor_pricing-8c01b7a7eda3 triggered at the same time.
Cost. Rollouts and Security Reviewer are included in the existing Teams ($40/user/mo) and Enterprise (custom) plans — no per-bot, per-PR, or per-finding surcharge on top of the seat price. The 10-day credits (50 changes for Teams, 500 for Enterprise) are an on-ramp for evaluation, not a usage cap. The hidden cost is observability: without Datadog/Grafana/Honeycomb (or an adapter) connected, Rollouts will not have signals to verify against and will not give you useful verdicts.
Risk — false confidence on "verified healthy." Rollouts compares post-deploy signals against the pre-deploy baseline. If your baseline is short, a regression that surfaces on day two is a "verified healthy" verdict on day one. Treat "verified healthy" as "within baseline noise on the signals Rollouts was configured to watch," not "no regression exists."
Risk — Security Reviewer dismissals are sticky per-PR. Dismissing a finding with a reason stops Security Reviewer from raising it again on that PR — meaning a security engineer who dismisses a finding on PR #4711 will not see it on the diff that re-introduces it in PR #4801. Use dismissals for genuinely out-of-scope findings; use the team-rules surface for cross-team policy.
Risk — gate at Teams/Enterprise only. The launch is explicitly not for Individual ($20/mo) or Hobby (free) plans. The cheapest path to either bot from Individual is a Teams upgrade at $40/user/mo — but that comes with the rest of the Teams feature set whether you want it or not.
Limitation — what I did not test. This is a documentation-comparison report. I have not enabled Rollouts or Security Reviewer against a real repo, not measured false-positive rate, not observed the monitoring-plan output on a real PR, and not tested the Security Reviewer dismiss-and-reraise behavior. Builders should plan a small proof-of-concept on a non-production repo before betting on either capability for a regulated workload.
Limitation — observability integrations are first-party partial. Datadog is named; "other telemetry providers" are documented but not enumerated. Cloudflare-native customers and AWS-only shops should ask Cursor Sales which telemetry adapters are supported in private beta vs roadmap.
Limitation — release-date precision. The Cursor blog post is dated Sep 23, 2026. The changelog entry did not surface in the source-watch signal until 2026-09-26 05:09:54 UTC; the launch is Sep 23, this article is Sep 26.
Rollouts and Security Reviewer are the first visible product of the Firetiger acquisition, and they are real product surfaces — not a press release in disguise. The monitoring-plan PR-comment shape is the substantive new piece: it forces instrumentation gaps to be visible at code-review time, lets the customer edit the plan, and gives a per-environment verdict rather than a binary green/red deploy hook. Security Reviewer's "reads code like a security engineer" framing is the right contrast with static analysis, and the dismiss-per-PR + team-rules combo gives a real policy surface.
The strategic read is more important than the feature read. Cursor is no longer a coding tool with adjacent features; it is a CI/CD-adjacent agent platform with a coding-tool front door. Three product surfaces shipped in the last four weeks (Self-Hosted Machines on Sep 2, Cursor Projects on Sep 12, Rollouts and Security Reviewer on Sep 23) all point in the same direction: build the path from writing code to running code in production, and make the agent the connective tissue. The Firetiger acquisition is the proof.
For builders, the practical takeaway is small and concrete: if you are on a Teams or Enterprise plan, enable both bots on a non-production repo this week. The monitoring-plan output is the artifact worth evaluating first — if your team's existing instrumentation is thin, the bot will tell you on day one. The 10-day credit window is the right time to find out.
For founders and VP Engineering, the bigger read is that Cursor is now a credible competitor to Datadog for incident context, to Sentry for production health, and to Snyk/Semgrep for security review — while still competing with Claude Code and Codex for the coding step. That is a lot of competitive surface for one product line, and it changes which vendor-consolidation conversations are worth having.
The honest caveats are the dismiss-sticky-per-PR surface and the gate at Teams/Enterprise. Both are real product decisions, not bugs, and a team that plans around them gets the full value; a team that discovers them in production gets a less pleasant onboarding.
Today / this week:
This month:
This quarter:
Skip if not in scope: