Atlassian's Code Reviewer is now sold with paid Jira

Code review is the slow part of software work. The pressure on it keeps rising. Atlassian says its own pull requests have grown from 46 lines to 102 lines in a year. Teams that lean on AI now wait twice as long for a human to look at the code.

Atlassian announced that Code Reviewer is generally available on 13 November 2026. The tool is an AI agent that reads a pull request. It leaves inline comments. It suggests fixes for bugs, logic errors, security issues and poor code structure. Atlassian includes it in paid Jira and Teamwork Collection seats. Usage is billed in Rovo credits.

Where it runs

Code Reviewer works today on Bitbucket Cloud and GitHub. Atlassian says GitLab support is next. The agent posts to the pull request as a bot. It can also check the diff against the Jira ticket's acceptance criteria. It can add a short summary when the change is large. A "Standards" beta lets a team write rules the agent must follow.

The numbers Atlassian quotes

Atlassian gives two figures from running it on its own repositories. Pull request cycle time dropped 31 percent. 39 percent of the agent's comments led to a code change. Both are Atlassian's own numbers on its own code. No third party confirms them.

Atlassian also reports a wait of 71 minutes for a human review on teams that lean on AI. The old wait was 32 minutes. Pull request size doubled in the past year. Those are the problem the agent is sold against. They do not measure the agent.

What this costs

Rovo credits are usage-based. A repository with many pull requests will cost more than a quiet one. A large pull request also costs more. Atlassian names no per-review price in the post. A team that already pays for Jira will see the credits added to its monthly bill. Set a cap before turning the agent on everyone.

What it means

For a team on Jira and Bitbucket, Code Reviewer is a feature of a tool they already pay for. Switching it on gives every pull request a first pass that catches easy problems. A missing null check. A dropped await. A secret in a config file. A human still reads the code and approves the merge. The agent shortens the first wait, while the final review remains the step that decides quality.

The risk is a team that treats the agent's silence as approval. Atlassian's own number cuts both ways. 39 percent of comments led to a change. 61 percent did not. A team that reads every comment is back to the same minutes it hoped to save. Set the Standards rules early. Check the Rovo credit burn after the first busy week.

Source

Meet Code Reviewer, now available with Jira, Atlassian.