Ultrareview code review, a deep code analysis feature, has been introduced by Anthropic and is available in Claude Code v2.1.86 and later. The feature runs on Claude Code’s web infrastructure and uses multiple reviewer agents to identify bugs in code branches and pull requests.
Ultrareview differs from the standard /review command by deploying a fleet of reviewer agents in a remote sandbox. Each finding is independently reproduced and verified, focusing on actual bugs rather than style suggestions. The multi-agent approach surfaces issues that a single-pass review might miss.
How Ultrareview Works
Users can launch Ultrareview from the CLI using the /ultrareview command. Without arguments, it reviews the diff between the current branch and the default branch, including uncommitted and staged changes. For GitHub pull requests, users pass the PR number as an argument, such as /ultrareview 1234.
The feature requires authentication with a Claude.ai account because it runs on Claude Code’s web infrastructure. Before launching, Claude Code displays a confirmation dialog showing the review scope, remaining free runs, and estimated cost. Reviews typically take 5 to 10 minutes and run as background tasks, allowing users to continue working in their session.
Pricing and Availability
Ultrareview is a premium feature that bills against extra usage rather than a plan’s included usage. Pro and Max subscribers receive three free runs through May 5, 2026. After the free allocation expires or is exhausted, each review costs approximately $5 to $20 depending on the size of the change.
The feature is not available when using Claude Code with Amazon Bedrock, Google Cloud Vertex AI, or Microsoft Foundry. Organizations that have enabled Zero Data Retention cannot access Ultrareview. Users must have extra usage enabled on their account or organization before launching a paid review.
Ultrareview Versus Standard Review
The standard /review command runs locally in a user’s session and completes in seconds to a few minutes, making it suitable for quick feedback during development. Ultrareview runs remotely in a cloud sandbox with a multi-agent fleet and independent verification, taking 5 to 10 minutes but providing deeper analysis.
Anthropic recommends using /review for fast feedback while iterating on code. Developers should use Ultrareview code review before merging substantial changes when seeking deeper analysis that catches issues a single review might miss. The feature represents an expansion of artificial intelligence capabilities in development workflows, leveraging code security verification through automated agent-based analysis.
Feature Status and Future Updates
Ultrareview is currently a research preview feature, and Anthropic stated that the feature, pricing, and availability may change based on feedback. Users can track running reviews using the /tasks command, which displays running and completed reviews, opens detail views, or stops reviews in progress. Stopping a review archives the cloud session, and partial findings are not returned.
The command runs only when users invoke it with /ultrareview; Claude does not start an Ultrareview on its own. Repositories that are too large to bundle prompt users to use PR mode instead, requiring users to push their branch and open a draft PR before running the command with the PR number.
Source: Claude Code Docs
