Claude Code Review Levels: Quick, Standard and Deep Explained

Last updated: October 20, 2026

Claude Code Review, Anthropic's managed pull request reviewer, now lets you pick how hard it looks. There are three review levels. Quick is a lighter, cheaper pass, Standard is the default, and Deep is the most thorough and most expensive. An organization Owner sets one level for the whole org and can override it per repository. The levels shipped in the Claude Code changelog for v2.1.304 (published October 19, 2026), alongside a new option to review every repository in a GitHub organization, including repositories added later.

Before this there was no depth setting at all. Now you can spend less on a routine internal repo and more on the code that pays for mistakes. If you are comparing reviewers, GitHub made the same move with Copilot code review effort levels, and our AI code review tools comparison covers the field.

What shipped#

From the Code Review setup docs:

Level	What it trades
Quick	A lighter pass. The fastest level, at the lowest cost
Standard	Balanced depth, speed, and cost. The default level
Deep	The most thorough pass. Takes the longest and costs the most

Org default: select a level under Review level and confirm with Change level. Every repository uses it unless it has its own.
Per repository: click Edit on the repository's row and pick a Review level. The first option, Default, follows the org. It saves as soon as you select it.
When it applies: a new level applies to reviews that start after the change.
No model picker: the Code Review overview says admin settings don't offer a model choice. The review level is the knob for thoroughness.

Both settings live in the Code review section of Admin settings > Claude Code on claude.ai, and only an Owner of the Claude organization can change them.

Which review level and trigger a Claude Code Review uses

%% title: Which review level and trigger a Claude Code Review uses
flowchart TD
  A["Pull request event"] --> B{"Repository Review runs setting"}
  B -->|"Once after PR creation"| C["Review on open, reopen, ready for review"]
  B -->|"After every push"| D["Review on each push"]
  B -->|"Manual"| E["Review only on @claude review"]
  C --> F{"Repository has its own level?"}
  D --> F
  E --> F
  F -->|"Yes"| G["Use repository level"]
  F -->|"No, Default"| H["Use organization level"]

How cost works#

There is no flat per-review price. Reviews are billed on token usage through usage credits, separately from your plan's included usage, and they are billed to the organization, not the developer who opened the PR. The cost of each review scales with four things: the review level, the size of the change, the complexity of the surrounding code, and how many candidate issues need verification.

The number of reviews matters as much as the level. Each repository has its own Review runs setting:

Once after PR creation: a review when a PR is opened, reopened or marked ready for review.
After every push: a review on each push. It catches new issues as the PR evolves and auto-resolves threads you fix, but it multiplies cost by the number of pushes. This is the default for a repository you add from the admin page.
Manual (@claude review): no automatic reviews. Comment @claude review for one review, or @claude review always to subscribe that PR to reviews on later pushes.

Automatic reviews skip drafts and pull requests from forks. A Deep review on every push of a busy repository is the most expensive combination available, so choose the level and the trigger together.

Pick a level per repository#

A sensible starting split:

Deep on the repositories where a missed bug is expensive: payments, auth, infrastructure-as-code, anything customer data flows through. Pair it with Once after PR creation so you pay for depth once per PR.
Standard as the org default, which is what you already had.
Quick on high-churn, low-risk repositories such as docs sites or internal tooling, where After every push keeps feedback flowing cheaply.

Then measure. The docs' own estimating method:

Set a monthly spend limit for the pilot.
Turn Code Review on for a few representative repositories with Review runs set to Once after PR creation or Manual.
After several reviews, read each repository's Average cost in the Repositories table.
Multiply by the PRs that repository opens in a month, and for After every push, by the typical pushes per PR.

A repository's average moves with its PR size and the level you choose, so re-run step 3 after you change levels. The analytics page shows PRs reviewed, spend this month, auto-resolved comments and a per-repository breakdown. Its costs are estimated at list price, so your Anthropic bill is the number to trust.

The other 2.1.304 change: cover a whole GitHub org#

You can now turn Code Review on for every repository in a GitHub organization, including repositories added later, by setting Review runs on the organization's row. It has three preconditions: the GitHub account you connected must own that GitHub org, the org must be on github.com (not GitHub Enterprise Server or a personal account), and a monthly spend limit must be set. The docs warn it "can raise spend significantly". Setting the org row applies its Review runs option to every repository in it, including ones you had configured individually, except repositories you turned off on their own rows. An organization row has no review level of its own, so each repository still uses its own level or the org default.

Gotchas and limits#

Spend limits stop reviews silently for the author. Code Review skips a review when the Code Review spend limit, the organization spend limit or the usage credits balance runs out, or when what is left is less than the expected cost of one review. Developers see a PR comment naming the cause, but nothing reviews that PR until someone raises the limit.

Levels are Owner-only. Developers can request reviews and change what gets flagged with REVIEW.md and CLAUDE.md, but they have to ask an Owner to change a repository's level or trigger. Decide who owns those requests before rollout.

Availability is narrower than Claude Code. Code Review needs a Team or Enterprise plan, GitHub hosting, and an org without Zero Data Retention, the HIPAA configuration or customer-managed encryption keys. Organizations that use Claude Code through Bedrock, Google Cloud or Foundry still need a Claude organization on one of those plans, and the reviews show up on the Anthropic bill.

A review never blocks a merge. Findings arrive as inline comments with a summary and a check run, but a review never approves or blocks the PR. If you want Deep findings to gate a merge, that rule has to live in your branch protection or your team's process.

The closest alternative#

GitHub's Copilot code review effort levels are the direct comparison: the same idea of trading depth for cost per review, inside GitHub's own billing. For teams already on Codex, the Codex code review plugin reviews from the agent rather than as a managed GitHub check. Claude Code Review's edge is that it runs on Anthropic infrastructure with no workflow files, and now has per-repository depth. Its cost is that pricing is usage-based with no flat per-review figure, so the pilot above is not optional.

What people are actually saying#

The change is less than two days old and we found no Hacker News or Reddit thread on review levels specifically. The broader argument that review capacity, not code generation, is now the constraint is the one we made in AI code review is the new bottleneck, and per-repository depth is a direct answer to it.
Until practitioner numbers exist, treat "Quick is cheaper, Deep is more thorough" as Anthropic's description, and get your own per-repository averages from the pilot.

What to do Monday#

Check your plan and data-retention settings against the availability list.
Set a Code Review spend limit at claude.ai/admin-settings/usage.
Pick one critical repository for Deep with Once after PR creation, and one busy low-risk repository for Quick.
After a week, compare Average cost across the two and adjust the org default.

FAQ#

What are the Claude Code Review levels?#

Quick is the lighter, fastest and cheapest pass. Standard is the default balance. Deep is the most thorough and costs the most. An Owner sets an org default and can override it per repository.

How much does a Claude Code Review cost?#

There is no flat price. Reviews bill on token usage through usage credits, and each review's cost scales with the level, the size of the change, the surrounding code's complexity and how many candidate issues need verification.

Who can change the review level?#

Only an Owner of the Claude organization, in Admin settings > Claude Code. Developers can request reviews with @claude review and tune what is flagged with REVIEW.md and CLAUDE.md.

Which plans include Claude Code Review?#

Team and Enterprise. It is not available with Zero Data Retention or the HIPAA configuration, or yet with customer-managed encryption keys.

Continue Reading#

GitHub Copilot Code Review Effort Levels Are GA: Lite vs Balanced - the same tradeoff on GitHub's side
Best AI Code Review Tools in 2026: CodeRabbit vs DeepSource vs Greptile Compared - how the rest of the field prices and scopes reviews
AI Code Review Is the New Bottleneck - why review depth is now worth paying for

Sources#