Claude Code review: how to set it up and make it catch more
· 9 minute read
Claude Code review is not one product. Anthropic ships a managed Code Review service for GitHub pull requests, a /code-review command you run in a local session, CI jobs through GitHub Actions or GitLab, and separate security review tools. This post explains those official options first, then how shared codebase context can help any of them catch more.
The facts below come from Anthropic's documentation, read on 22 November 2026 (Asia/Calcutta). Graphify is not a replacement for any of these tools. It is shared context (callers, tests, and impact) that a review can use.
The official ways to get a Claude Code reviewLink to The official ways to get a Claude Code review
Anthropic documents several ways to have Claude review code. They are not interchangeable.
| Option | Where it runs | Who can use it | How it starts |
|---|---|---|---|
| Code Review (managed service) | Anthropic infrastructure, comments on GitHub pull requests | Team and Enterprise plans, after an Owner turns it on | A repository setting, or an @claude review comment |
/code-review (alias /review) | Your local Claude Code session | Any plan | You type the command in the session |
Ultrareview (/code-review ultra) | A cloud sandbox, started from Claude Code | Requires a claude.ai sign-in; availability varies | You run /code-review ultra |
| Claude in your own CI | Your GitHub Actions or GitLab CI runners | Anyone with a Claude API key or cloud provider credentials | Your pipeline events, such as each pull request or merge request |
Security review is covered later in this post. It is a related set of tools, not a fourth name for the same feature.
Code Review on GitHub pull requestsLink to Code Review on GitHub pull requests
The managed Code Review service reviews pull requests in your GitHub repositories, including GitHub Enterprise Server and GitHub Enterprise Cloud with data residency when those are set up. Several agents read each change alongside the surrounding code, looking for issues such as logic errors, security vulnerabilities, or broken edge cases. A verification step checks findings against the code before they are reported.
Findings arrive as inline comments, ranked by severity, with a summary and a check run named Claude Code Review. The check run does not conclude as a failure. A review never approves or blocks the pull request, so the decision to merge stays with your team. By default, Code Review focuses on correctness: bugs that would break production, not formatting preferences or missing test coverage. You can change what it flags by adding guidance files, including REVIEW.md, to the repository.
To turn it on, an Owner of your Claude organization follows Anthropic's setup guide: connect GitHub, install the Claude GitHub App, choose repositories, and set when reviews run. Each repository can review once when a pull request opens, after every push, or only on request. An Owner also chooses a review level for the organization (Quick, Standard, or Deep) and can give a repository its own level.
Whichever setting you use, someone with write access can comment @claude review on a pull request to start a review. Official details are in Anthropic's Code Review documentation.
/code-review in a local sessionLink to /code-review in a local session
/code-review is a command you type at the Claude Code prompt. Claude reviews the changes on your branch, looks for correctness bugs, and reports what it found in the conversation. /review is an alias of the same command.
The command works on any plan. The review runs in your session, so it counts toward that session's usage the same way the rest of the session does. With no target, it reviews your branch's commits ahead of its upstream branch plus your uncommitted changes. If the branch has no upstream, Claude compares against main, or against the previous commit. You can also point it at a range, a path, or a pull request number:
/code-review
/code-review main...my-feature
/code-review src/auth
/code-review 1234The local review follows your CLAUDE.md. It does not read REVIEW.md. That file configures the managed Code Review service on GitHub pull requests, not this command.
/code-review ultra starts Ultrareview, a deeper multi-agent review in a cloud sandbox. It is a mode of the same command, not a separate product you install. Anthropic's local /code-review docs and Ultrareview docs cover targets, effort levels, and limits.
Claude Code review in GitHub ActionsLink to Claude Code review in GitHub Actions
If you want Claude to review in your own pipeline rather than on Anthropic's managed service, you can run Claude Code in GitHub Actions or GitLab CI/CD. The review then runs on your CI runners, triggered by your pipeline's events, such as every pull request or merge request. This path needs a Claude API key or cloud provider credentials. It is the option to use when the managed Code Review service is not available on your plan, or when you want the job to live next to your other CI checks.
Security review in Claude CodeLink to Security review in Claude Code
Anthropic also ships security-specific review tools. They are distinct from Code Review and from /code-review.
The official security-guidance plugin reviews the code Claude itself writes, in the same session, and can fix what it finds. Install it from the official marketplace:
/plugin install security-guidance@claude-plugins-officialOnce installed, it runs automatically. There is no separate command to remember. It is meant to catch issues such as injection, unsafe deserialization, and unsafe DOM APIs before the code reaches a pull request.
/security-review is different. It is an on-demand, single pass over the changes on your current branch, run when you ask. It does not replace the plugin, and it does not scan a whole repository.
For a deeper scan of a repository or of a chosen set of changes (a branch diff, a pull request diff, or a single commit), install the Claude Security plugin:
/plugin install claude-security@claude-plugins-officialThat plugin is a multi-agent vulnerability scan you start in a session. It is distinct from the security-guidance plugin, from /security-review, and from the managed Code Review service.
How shared codebase context improves a reviewLink to How shared codebase context improves a review
The tools above already have a view of the change: the diff, the surrounding file, and, in a local session, whatever Claude reads next. What they do not automatically share with each other, or with the rest of your team, is a durable map of how the code connects.
A review gets sharper when it can ask three questions that the diff alone often cannot answer:
- Callers. Who still calls the function you changed, including callers that do not share a name with it?
- Tests. Which tests reach that code, so a missing test is a fact rather than a guess?
- Impact. What else sits on the path of the change, across files or packages?
That is shared codebase context: a code graph of calls, imports, and references, plus the decisions, incidents, and lessons your team has already recorded. Our post on graph-based review versus diff-only review spells out the difference.
Graphify builds that graph for the repositories you connect and keeps it current as the default branch moves. In Claude Code, the hosted MCP server can answer those caller, test, and impact questions while you run /code-review, or while you read findings from the managed Code Review service. The graph does not replace Anthropic's reviewer. It gives the reviewer, and you, a way to check a finding against real callers and tests instead of against the hunk alone.
If you are comparing reviewers more broadly, our AI code review tools comparison sits next to this guide. For the Codex side of the same problem, see Codex code review.
Try itLink to Try it
Start with Anthropic's own path. If your organization is on a Team or Enterprise plan, have an Owner follow the Code Review setup guide. If you want a review in the session you are already in, run /code-review. For security, install the official security-guidance plugin and use /security-review when you want a pass on the current branch.
To give those reviews shared codebase context, connect a repository at app.graphify.com and add the hosted MCP server in Claude Code. The Claude Code integration and the reviews docs cover the setup. Prefer to stay local? Run uv tool install graphifyy for the Graphify CLI. The Graphify CLI is licensed under Apache 2.0.
Related reading: AI PR review grounded in a code graph, how to reduce AI code review noise, and Codex code review. Sources, read on 22 November 2026 (Asia/Calcutta): Anthropic's Code Review, Review locally with /code-review, Ultrareview, GitHub Actions, security-guidance, and Claude Security plugin pages.
FAQLink to FAQ
What is Claude Code review?Link to What is Claude Code review?
Claude Code review refers to several official Anthropic options. The managed Code Review service comments on GitHub pull requests for Team and Enterprise plans after an Owner turns it on. The /code-review command, also available as /review, runs in your local Claude Code session on any plan. You can also run Claude in GitHub Actions or GitLab CI, and use separate security review tools.
How do I run /code-review in Claude Code?Link to How do I run /code-review in Claude Code?
In a Claude Code session, type /code-review at the prompt. With no target, Claude reviews your branch's commits ahead of its upstream branch plus your uncommitted changes. You can also point it at a path, a commit range, or a pull request number, for example /code-review 1234. /review is an alias of the same command.
Does the Code Review feature work on every plan?Link to Does the Code Review feature work on every plan?
No. The managed Code Review service that comments on GitHub pull requests is for organizations on a Team or Enterprise plan, and an Owner has to turn it on and connect GitHub. The /code-review command in a local Claude Code session works on any plan and counts toward that session's usage.
How do I request a review with @claude review?Link to How do I request a review with @claude review?
After an Owner has turned on the managed Code Review service, someone with write access to the repository can comment @claude review on a GitHub pull request to start a review. Each repository can also be set to review when a pull request opens, after every push, or only on request.
Is /review the same as /code-review?Link to Is /review the same as /code-review?
Yes. Anthropic's docs state that /review is an alias of /code-review. Both run the same local review in your Claude Code session.
Can Claude Code do a security review?Link to Can Claude Code do a security review?
Yes. The official security-guidance plugin reviews the code Claude itself writes, automatically, once you install it. /security-review runs a single security pass on the current branch when you ask. The Claude Security plugin is a deeper on-demand scan of a repository or a set of changes, and is distinct from those two.