Agents in Jira, in short: Atlassian’s feature that lets teams assign Jira work items to AI agents, @mention them in comments and trigger them from workflow transitions went into open beta on February 25, 2026. By May 2026 it was generally available for Jira Cloud Standard, Premium and Enterprise plans, according to Atlassian’s own update. It is no longer a beta, and it is metered: Atlassian’s support page says each agent run uses Rovo credits, with credit allowances applied starting December 3, 2026.
The agents can be Rovo agents built by Atlassian or third-party agents such as the GitHub Copilot coding agent. The pitch, in Atlassian’s words, is to do “10x the work, without 10x the chaos,” by keeping people and agents on the same work item, inside Jira’s existing permissions and audit trail. This page was first written during the beta and has been updated on October 13, 2026 to reflect where the feature stands now.
What are AI agents in Jira and how do you use them?
Atlassian’s support documentation lists four ways to involve an agent in a work item:
- Agent picker or assignee: open a work item, choose an agent, and it works from the summary and description (it may also use comments and attachments).
- @mention: type @ and the agent’s name in a comment.
- Workflow transition: add a “Trigger agent action” rule, with an optional custom prompt, so the agent runs whenever anyone moves an item into that status. This needs the Space admin role and the Edit workflows permission.
- Board column: in team-managed spaces only, add an agent to a column and it runs when work moves in.
Setup is a one-time step per agent: turn on the Work item toggle under Configuration, then Surfaces, and make sure Rovo is enabled for the site. Output is private by design. It is visible only to the person who assigned the agent (or triggered it) until they choose to share it, for example with the Draft comment option. That keeps a person in the loop before anything reaches the team.
What has happened since the February 25 announcement
- February 25, 2026: Atlassian published its announcement and called the feature an open beta. TechCrunch reported the same day that Tamar Yehoshua, Atlassian’s chief product and AI officer, framed the point as coordinating humans and agents in one place.
- February 16 to 23, 2026 changelog: Atlassian’s cloud changelog marked the rollout as an Open Beta and said it was not yet available for Jira Product Discovery, Jira Service Management or Jira Service Desk. It listed the GitHub Copilot coding agent as coming soon.
- Partner agents: the announcement names Amplitude, Box, Canva, Figma, GitHub Copilot, HubSpot and Intercom among the providers of agent skills available through Atlassian’s MCP server gallery.
- May 2026: Atlassian’s community post was updated to say agents in Jira is generally available for all Jira Cloud Standard, Premium and Enterprise plans. The post does not give a day in May.
- October 13, 2026 (today): the support documentation describes agents in Jira as a premium AI feature that consumes Rovo credits.
What do agents in Jira cost?
Per Atlassian’s support page, each time an agent is invoked by assignment, @mention or a workflow or column trigger, credits come out of the organization’s Rovo credit balance. The number of credits depends on the agent’s configuration and the thinking mode it uses. Allowances apply starting December 3, 2026.
If an organization goes over its allowance, what happens depends on its plan. With extra usage enabled, agent orchestration continues without interruption, admins are notified at 80% and 100% of usage, and the extra usage is charged on the next invoice. Organizations can also buy usage packs. Admins can track consumption in Atlassian Administration under Insights, then Platform usage. We did not find a published per-run credit figure on these pages, so the cost of a given agent is something to measure during a pilot.
What it means for a small team
Atlassian’s general availability note names the Standard, Premium and Enterprise plans. It does not mention the Free plan, and we did not find plan details beyond that, so a team on Free should check its own Jira admin settings before planning around the feature. For teams on a paid plan, the practical questions are about credits and review, not about whether the feature exists.
Our read: agents in Jira are most useful as a first-draft helper for recurring, low-risk work such as drafting a changelog, summarizing a long ticket thread or preparing a customer update, where one person reads the draft before it goes anywhere. Because output stays private until shared and agents run inside existing permissions, the person who owns the ticket stays in charge. That is the version that gives a small team time back without taking decisions away from it. Start with one repeating task, record the credits it uses over two weeks, and decide from that number. For wider context on picking tools, see our guides to AI automation for small business, AI software for small business and AI for small business.
What we could not confirm
We could not confirm from Atlassian’s pages whether agents have since been added to Jira Product Discovery or Jira Service Management, which were excluded during the February beta, or what a typical agent run costs in credits. Third-party reports of additional partner agents (for example, other coding tools) were not on an Atlassian page we could read, so we have left them out.
