Introducing Agentic Pipelines in Bitbucket (open beta)

[#1]

Hi everyone!

Today we’re opening up a new way to use Bitbucket Pipelines: Agentic Pipelines, AI automation for all the dev chores you don’t want to do.

If you’re like most teams, a surprising amount of your week goes into the “work around the work”: keeping READMEs and docs up to date, cleaning up feature flags, triaging security issues, writing release notes, or trying to make sense of a big PR before you review it. It all matters, but it’s repetitive, easy to defer, and painful when it piles up.

Agentic Pipelines lets you delegate those chores to AI, directly from Bitbucket Pipelines. You define an agent once in your bitbucket-pipelines.yml, give it clear constraints, and then let it run on a schedule or on events like PR creation/merge. The agent can then do things like:

• Propose README and docs updates on every merged PR

• Read a Snyk report and raise a PR with suggested security fixes

• Clean up feature flags after a feature is rolled out

• Generate high‑quality release notes from your commit history

• Summarise large PRs and post the summary as a comment

• Identify low‑coverage areas and raise PRs with additional tests

All changes go through a standard Bitbucket PR, with scopes you configure explicitly, so you stay in control.

Agentic Pipelines is now in open beta. You don’t need to be an existing Pipelines user – you just need a Bitbucket Cloud repo and a Rovo Dev license to get started. We’ve published public example repos for each of the use cases above so you can copy, tweak, and run them in your own projects.

Using other AI agents? Stay tuned! Support for third-party CLIs like Claude Code are coming very soon.

👉 Learn more and see the examples: Announcement blog

We’re already seeing Atlassian engineering teams get hours back every week by automating this kind of work. I’m keen to see what you build with it – especially the unexpected use cases we haven’t thought of yet.

Try it out and send us your your feedback: what works, what doesn’t, and what you’d like to automate next.