Claude Code /ultrareview: Multi-Agent Deep Review for Critical PRs

After fixing code, there's always a nagging worry. "Did I break something else?" Especially for parts where mistakes are costly, like payments or login. For times like these, Claude Code (an AI coding tool used in the terminal) has a feature where multiple AIs look at the same code at once, deeply. That was /ultrareview, and now using /code-review ultra is recommended.

Definition (What It Is)

/ultrareview is a feature where multiple agents review code deeply in a cloud sandbox.

• Cloud sandbox: an isolated external environment, not your computer, where code can be run safely. You can verify it without affecting your PC.
• Agent: an AI worker that works on its own. When multiple agents view the code from different perspectives (security, performance, bugs, etc.), there are fewer gaps than when one person looks alone.
• PR (Pull Request): a bundle of changes you submit in Git (a tool for managing code changes) saying "please merge these changes."

Now, instead of a separate command, attaching the ultra option to /code-review is recommended.

How to Use It (By Difficulty)

Basic — Run a Precise Review

/code-review ultra

Multiple agents review the changes in depth. It takes more time and cost than a regular review, but it examines things that much more thoroughly.

Common Pitfalls (When Applicable)

• A precise review is a heavy task that uses a lot of resources. Running it every time even for trivial changes wastes time and cost. The key is to use it selectively, only for truly important changes.
• If you still remember the old name /ultrareview, in the current version it's correct to use /code-review ultra. Just keep in mind that the name has changed.

Real-World Example

There was a PR that changed payment logic. Since it's a part where mistakes leak money, a regular review alone felt uneasy. Having multiple agents review it precisely with /code-review ultra pointed out a missing exception handling that a person would likely have overlooked, reducing the risk before merging.

Taking It Further

You don't need to precisely review every change. It's efficient to "be selective and focused": normally review quickly with a light /code-review, and turn on ultra only for high-cost-of-error changes like payments, authentication, or data migration.

Summary

/code-review ultra (formerly /ultrareview) is an advanced review feature where multiple agents review code deeply in the cloud. Since it's heavy and costly, using it sparingly only for truly important changes makes it a reliable safeguard that heads off big incidents in advance.

Based on: Claude Code v2.1.154 (2026.05)