Atlassian adds Jira features for always-on AI agents

Atlassian frames the next phase of software development as always-on agentic AI, where agents run for hours or days without iterative prompting and pick up work when needed. That shift is why the company says supervision must change: letting agents work from start to finish with less oversight brings caveats around trust, grounding, shared context, communication, institutional memory and validation.

The new features are organized around that problem. Code Context, built on Atlassian's Teamwork Graph, aims to give agents a view of project architecture, decisions and standards, which the company says they currently lack. Agent Space Settings and Agent Context Controls then govern where agents operate, what they can see and what they can do, similar to how teams control employee access. On the execution side, AutoDev turns backlog work into merge requests in Jira, a standards system conforms code, a review agent flags problems, and a DevDocs agent keeps technical documentation in Confluence current directly from code repositories.

Atlassian pairs this with accountability tools, since it says there is no perfect playbook and best practices are still being built. Whether the approach holds up likely hinges on whether grounding agents in project architecture and standards actually reduces the failures the company describes, and on whether audit logs and the agent usage dashboard give teams enough visibility into spend and outcomes. For enterprise engineering organizations, the appeal of moving from foreground coding tools to always-on agents depends on that governance proving workable at scale.