News

Atlassian Launches Code Reviewer, an AI Agent for Pull Request Review, Now Generally Available

· Atlassian Blog

Atlassian has made Code Reviewer generally available to paid Jira and Teamwork Collection customers. The AI agent reviews pull requests in Bitbucket Cloud and GitHub, with GitLab planned, and uses Jira and Teamwork Graph context to check changes against linked work item acceptance criteria. Atlassian cites data showing heavy AI-coding teams wait 71 minutes for human review, up from 32, while median pull request size rose from 46 to 102 lines. Internally, the company reports a 31% drop in median cycle time, and 39% of comments led to code changes. A Standards feature is in beta, and usage consumes Rovo credits.

Atlassian has announced the general availability of Code Reviewer, an AI agent designed to give every pull request a quick first-pass review. It is included with paid Jira and Teamwork Collection subscriptions, and it targets what Atlassian describes as a growing bottleneck: as AI coding agents help developers produce more code, human reviewers are struggling to keep pace.

The problem: review as the new constraint

Atlassian points to findings from its Q3 State of AI Impact in Engineering Report, produced with DX. According to that report, teams that lean heavily on AI to write code wait more than twice as long for a human reviewer, a median of 71 minutes compared with 32 previously. Median pull request size has also more than doubled over the past year, rising from 46 to 102 lines of code. The pitch is that faster code generation shifts pressure downstream, so review capacity needs to scale as well.

What Code Reviewer does

The agent works on pull requests in Bitbucket Cloud and GitHub, with GitLab support listed as coming soon. It leaves inline comments that flag problems and propose fixes that authors can apply directly within the pull request. The categories covered include bugs, logic errors, security risks, and maintainability concerns. For larger changes it also generates a summary to orient human reviewers.

A distinguishing feature is the link to Jira and the Teamwork Graph. Code Reviewer compares a pull request against the acceptance criteria in linked Jira work items, so the review reflects the intent behind the change rather than only the code. A built-in quality filter suppresses low-value comments, aiming to keep output useful instead of noisy.

Internal results

Atlassian reports that since deploying the tool on its own code, median pull request cycle time fell by 31%, and 39% of the agent's comments led directly to a code change in the following commit. These are vendor-reported internal figures rather than independent benchmarks, but the comment-acceptance rate is a useful signal of review relevance.

Standards (beta)

Atlassian is also previewing Standards, a beta capability that lets organizations document coding standards and attach them to specific repositories. Code Reviewer then evaluates each pull request against the standards for that repository, helping enforce consistency regardless of which human reviewer is involved. Interested teams can join a waitlist.

Security and administrative controls

Each review runs in a temporary, isolated environment that is deleted once the review completes. Atlassian says code is not retained and is not used to train models. Administrators decide which workspaces and repositories get automatic reviews and can revert them to manual mode or disable the tool altogether.

Setup and best practices

Code Reviewer is enabled by default for eligible sites but in manual mode, meaning it only reviews when a pull request author requests it. To get the full benefit, admins can switch on automatic review. In Bitbucket Cloud, workspace admins go to Workspace settings, then AI, then Code Reviewer, and change the trigger to every new pull request or every commit, optionally setting it as the workspace default. For GitHub, repositories are managed through Jira, and Jira site admins configure the trigger under Developer AI reviewer settings, applying it across an organization or tuning individual repositories. GitHub users must first install the GitHub for Atlassian app, and Bitbucket users must connect their workspace to Jira.

Atlassian advises two habits for better results: include the Jira work item key in branch names or pull request titles so the agent can find the linked issue, and write specific, testable acceptance criteria rather than vague ones.

Pricing

The feature is bundled with eligible subscriptions, but it consumes Rovo credits under Atlassian's usage-based pricing policy, so heavy automatic review could generate additional cost.

Industry implications

The launch positions Atlassian in the crowded AI code review segment alongside tools from GitHub, standalone review startups, and coding-agent vendors. Its main differentiator is deep integration with project-management context in Jira, which competitors focused purely on code diffs may lack. The move also reflects a broader trend: as code generation becomes cheap, vendors are racing to automate validation, review, and governance, and to anchor those workflows in existing enterprise platforms. Support for both GitHub and Bitbucket is strategically important, since it extends Atlassian's reach beyond its own repository hosting.

Referenced links