Atlassian Plans Jira Controls for Long-Running AI Coding Agents

Atlassian is preparing Jira for coding agents that can work in the background for hours or days instead of waiting for repeated prompts. Its upcoming capabilities connect backlog automation with access controls, standards-based review and records of each agent run.

From a ticket to a proposed change

The package is designed to help engineering teams run and govern AI coding agents across the software-development lifecycle. AutoDev scans backlogs for work and turns items into code merge requests inside Jira. A standards system and dedicated review agent then check those merge requests against defined standards and flag problems.

Before an agent acts, Code Context is meant to give it a view across complex codebases that span multiple repositories. It indexes codebases into Atlassian’s Teamwork Graph, where agents can query connected repositories and draw on signals from Jira, Confluence, Loom and third-party connectors. Atlassian says the system uses both lexical search for exact matches and semantic search for natural-language requests.

Atlassian has made Code Context available through the Teamwork Graph command-line interface and says coding agents can retrieve relevant context before acting. The feature is in early access and gradually rolling out through open beta. It supports Bitbucket and GitHub repositories.

Access and accountability travel with the agent

Agent Space Settings and Agent Context Controls are designed to set where agents operate, what they can see and what actions they can take. For Code Context, Atlassian says administrators must opt in before indexing begins, and users and authorized agents can retrieve only code they already have permission to access through source control.

What teams can inspect

• Each agent run generates audit logs and diagnostics covering throughput, quality, adoption and cost.
• An AI agent usage dashboard shows who uses the tools, how they are used and team-level outcomes.
• A DevDocs agent generates or updates technical documentation in Confluence from code repositories.

A context result, not a workflow verdict

Atlassian reports that agents enriched by the Teamwork Graph produced 44% more accurate results while using 48% fewer tokens than agents without it in internal benchmarks. Those figures concern the context layer, not the full upcoming Jira package in customer use. The new controls describe how teams could constrain and inspect agent activity; they do not show how long-running automation will perform after deployment.