Bitbucket's AI code review is Cloud-only, leaving Data Center teams to build their own

What the native feature is

Bitbucket's built-in AI code review is sold under the name "AI-assisted code review" and runs on Rovo Dev, Atlassian's agent for software work. According to a dev.to write-up by Emil Reiter that cross-checked Atlassian's own pages on 17 September 2026, Rovo Dev handles code planning, code generation, review passes and a share of the repetitive grunt work — but only on Atlassian's hosted platform.

The plan list contains only Cloud tiers

The decisive evidence sits in the FAQ on Atlassian's Bitbucket AI page. Rovo is offered to customers on Standard, Premium or Enterprise editions of Bitbucket, Jira, Confluence, Jira Service Management or the Teamwork Collection — every one of them a Cloud subscription. No Data Center or Server tier appears anywhere in that list, which formally scopes the native reviewer to the cloud product.

Silence in the Data Center docs

The dev.to write-up also searched Atlassian's Bitbucket Server documentation, the docs tree covering Data Center, for the feature's name and got no matches. The author adds the necessary caveat: missing documentation is not proof that a capability will never ship. But combined with a published plan list that contains only Cloud tiers, and with nothing written for the self-managed edition, the working conclusion for administrators is that AI review on Data Center is a bring-your-own exercise as of 17 September 2026.

The seams for a third-party reviewer

Integration is therefore the only route. The write-up identifies two supported hook points on Data Center: pull-request triggers and build pipelines. Atlassian documents repository hooks and branch permissions for the self-managed edition, and the product ships pre-receive hooks that remain installed but switched off until an administrator enables them per project or per repository. An external review service, or a CI job that calls a model and posts its findings back to the pull request, attaches at exactly those seams.

The merge gate is yours to build

Wiring a reviewer in is only half the job. On Data Center, merge control belongs to branch permissions, usually paired with required builds and the pre-receive hooks. A team that wants a hard rule — nothing merges without an AI pass — has to assemble it from those primitives itself. Whether the reviewer's verdict can genuinely block a merge or only shows up as an advisory comment is a configuration decision Atlassian leaves to each customer.

Why it matters

The write-up crystallises a split many teams discover only mid-evaluation: Atlassian's flagship AI capability answers the Cloud question and stays silent on the Data Center one. Shops running self-hosted Bitbucket deliberately, whether for data residency, regulation or scale, are left with two options — pay for a third-party reviewer and engineer the merge gate around it, or reconsider a move to Cloud to get Rovo Dev natively.

It also fits a broader pattern. Vendors are landing marquee AI features cloud-first, because that is where usage can be metered and model access controlled, while self-hosted editions keep their older guarantees but stop being where new AI functionality debuts. For Data Center teams, the current documentation should be read as a deprecation of expectations rather than a temporary gap: as of the checked date, the plan is third-party tooling behind an enforcement layer you configure yourself.