On April 16, 2026, Claude Code rolled out /ultrareview as a research preview alongside Claude Opus 4.7. Unlike the local /review command, this feature launches a set of independent agents inside Anthropic's remote sandbox. These agents analyze the diff between your current branch and the default branch in parallel, taking 5 to 10 minutes before returning results as notifications to your session.
How /ultrareview Differs from /review
/review runs locally and offers near-instant feedback, suitable for quick checks while coding. However, its suggestions often mix code style and naming convention feedback with genuine bug alerts, making it hard to prioritize.
/ultrareview verifies each finding via independent agents. Anthropic says it targets real bugs, not style advice. Security, correctness, architecture, consistency, and test coverage are each handled by a separate agent, then aggregated. The official use case is deep review before merging critical changes, especially for auth logic and data migration—areas where errors have wide impact and high debug cost. For minor frontend tweaks or console logs, /review suffices.
Usage and Pricing: Free Tiers, Then Pay Per Run
Two ways to start: type /ultrareview in the CLI (it captures the diff including uncommitted changes), or specify a GitHub PR number like /ultrareview --pr 123. Results come later as notifications.
Requires Claude Code v2.1.86 or higher and a logged-in Claude.ai account.
Pricing is extra usage, not deducted from your subscription plan. Pro and Max users get 3 free runs before May 5, 2026, one-time only. After that, each run costs $5 to $20, depending on diff size—more changes, higher cost. Team and Enterprise users have no free runs, charged from the first use.
Supports not available for: Claude Code deployed on Amazon Bedrock, Google Cloud Vertex AI, Microsoft Foundry; organizations with Zero Data Retention (ZDR) enabled; and Team/Enterprise accounts (no free tier). Enterprise users often hit the ZDR wall—many companies keep ZDR on, and engineers may not realize they can't run /ultrareview until they try. Check your org's data policy in Claude.ai settings or ask your IT admin.
FAQ
Q1. Can /ultrareview and /review run together?
A: Yes, but no need. Use /review during coding for quick feedback, then /ultrareview before merging for deep scrutiny. They don't conflict, but /ultrareview costs money, so don't trigger it for every small change.
Q2. When do the 3 free uses start counting?
A: From the first successful /ultrareview execution, not from your subscription upgrade date. Deadline is May 5, 2026. Unused free runs expire after that, and you'll be billed $5–$20 per run. The 3 uses are one-time, not refreshed monthly.
Q3. Can a Team account use /ultrareview?
A: Yes, but each run costs $5–$20, no free tier. If your org has ZDR enabled, /ultrareview won't work—that's a hard limit, not a billing issue.