Claude Code Review leaves preview with adjustable review depth

Three levels of review

Anthropic moved Code Review out of research preview on October 15 for eligible Claude Team and Enterprise organizations. The service examines proposed code changes in GitHub pull requests and posts findings beside the relevant lines.

Administrators can choose Quick, Standard, or Deep for each repository; Standard is the default. Anthropic says most reviews now take 6–10 minutes, compared with roughly 20 during preview. Those timings are vendor claims, not our measurements.

Source 1Source 2

Access requires organization setup

A Claude organization Owner must enable the service and arrange GitHub App access to the selected repositories. GitHub Enterprise Server and Enterprise Cloud with data residency are supported. Organizations using Zero Data Retention, HIPAA configuration, or customer-managed encryption keys cannot use it.

Source 2Source 4

Frequency matters as much as depth

Reviews consume separately billed usage credits, with no flat price per review. Cost varies with review depth, change size, surrounding-code complexity, and verification work. Newly added individual repositories default to reviews after every push. Administrators can instead choose reviews when a pull request opens or only on request, and set a monthly spending limit.

Our reading: the useful change is finer control over where teams spend review effort. A frequently revised pull request can generate several charges, so choosing a lighter review level alone does not control total spending.

The service reads code without building or executing it. It neither approves nor blocks merging, and a clean report can still miss bugs. Its practical role is an additional review layer alongside tests and human judgment.

Source 2Source 3Source 4