Atlassian adds Jira controls for AI software agents

Atlassian has announced a set of Jira and DX features aimed at teams using AI agents in software engineering, with the company positioning the release as a way to bring agent activity into existing planning, coding, review and measurement processes.

According to Atlassian, the tools are intended to give engineering and platform teams more control over what context AI agents can access, what tasks they can perform and how their output is assessed. Several elements remain in phased release, including open beta, private early access and staged general availability.

A key feature is Code Context, built on Atlassian’s Teamwork Graph. The company says it will allow Rovo and coding agents to draw on information across multi-repository codebases, including architecture, requirements and project data. Atlassian also plans Agent Context Controls, which it says will let platform teams restrict the Jira and Confluence spaces available to agents.

The company is also adding agent loops in Jira, designed to identify well-defined, unassigned backlog tasks, send them to the Jira Coding Agent, and open pull requests for human review. Other planned controls include Standards, for defining coding rules linked to repositories, and AI review, which uses an agent to assess pull requests against organisational standards before code is merged. Final approval is described as remaining with developers.

Atlassian is extending its DX product with DX for Agentic Development, which it says will measure AI use across throughput, quality, adoption and cost. A Jira Agent Usage Dashboard is also planned to show which agents are appearing in workflows and how sessions relate to Jira work items.

Atlassian cited its 2026 AI SDLC study as finding 94% of engineering leaders use AI, while 6% have systems to scale it across the full development lifecycle. The methodology and sample details were not included in the supplied material.

Code Context is described as being in open beta for paid customers. Agent loops, Standards and AI review are in private early access. Other availability timings remain broad, including releases expected in the coming months or this quarter.