Atlassian brings AI agents to Jira with MCP support

Atlassian has launched an open beta of AI agents in Jira, bringing automated work execution into the interface many teams already use to plan and track tasks. It also announced new releases tied to Model Context Protocol (MCP), an emerging standard for connecting AI agents to software tools and data sources.

The updates aim to keep agent activity inside established work management processes, rather than in separate chat threads or standalone pilots. Atlassian positioned Jira as a system for coordinating work across people, software tools, and AI agents.

Agents in Jira lets teams assign tasks to Atlassian's Rovo agents as well as third-party agents that support MCP. Users can also pull agents into Jira comments for back-and-forth exchanges. Agents can also be embedded in workflows, enabling them to take actions and update items directly within Jira processes.

Workflow controls

Agents running in Jira operate within Jira's existing structures, including project configurations, permissions, audit trails, and approval flows. This addresses a common concern for larger organisations, where automation can create new compliance and oversight requirements if actions occur outside controlled systems.

Jira is widely used by software teams and business functions that run structured work through tickets and queues. Adding agent capabilities inside Jira aligns with a broader enterprise shift, as vendors introduce autonomous or semi-autonomous AI features into systems of record.

Tamar Yehoshua, Atlassian's Chief Product and AI Officer, said the release is designed to reduce coordination risks when organisations deploy multiple agents and tools at once.

"Work is changing fast: people are now orchestrating across agents, tools, and cross-functional teams. Without clear coordination that can easily turn into chaos," said Yehoshua. "We're focused on helping teams turn that complexity into real productivity. With these new capabilities, we're bringing agents into the tools and workflows customers already love and trust, and giving them an open, governed way to make those agents part of the team at enterprise scale."

MCP investments

Alongside Agents in Jira, Atlassian outlined new investments in MCP, which is intended to provide a common way for AI agents to access tools, data, and workflows. It is adding more MCP-related connectors and expanding how Rovo interacts with third-party applications.

Adoption data from larger customers points to growing enterprise use of MCP services tied to Rovo. Enterprises account for nearly 50% of all Rovo MCP Server usage, and customers on paid Atlassian editions account for 93% of usage.

Atlassian announced two product releases under its MCP programme. The first adds MCP skills to Rovo, allowing Rovo agents to connect to MCP-enabled third-party applications. Examples include Amplitude, Box, Canva, Figma, and Intercom. The goal is for agents to retrieve context and take actions across tools while staying connected to Jira and Confluence activity.

The second release is the general availability of Rovo MCP Server, an Atlassian-hosted MCP server that provides MCP-compatible AI clients with a single connection into Jira and Confluence. Compatible clients include Claude by Anthropic, Cursor, Google's Gemini CLI, Lovable, and WRITER.

Open ecosystem

The announcements reflect a bet that customers will adopt multiple AI agents rather than rely on a single assistant. That raises questions about identity, access control, auditability, and how teams trace decisions back to source data and approvals. Atlassian's emphasis on permissions and audit trails suggests it wants agent actions to remain within familiar governance models, rather than creating a parallel layer of automation.

For customers, the Jira open beta offers a new way to operationalise AI work alongside human work. It also creates a clearer record of what an agent did, when it did it, and which workflow rules applied-important in regulated settings and change-controlled environments, where automated actions can create risk if they are hard to track.

The MCP work also puts Atlassian among a growing group of vendors and standards efforts aimed at making agent integrations less bespoke. If MCP adoption increases, organisations may find it easier to swap agent providers or run different agents for different tasks while keeping a consistent integration model with tools such as Jira and Confluence.

The open beta of Agents in Jira and the new MCP releases extend Atlassian's push to place AI features inside day-to-day work systems, with Rovo and third-party agents operating under the same project structures and controls teams already apply to human work.