Atlassian Inc. is moving artificial intelligence agents directly into its Jira project management and issue-tracking software, calling it a natural evolution of enterprise collaboration in the AI era.

The company today announced an open beta test of a feature that allows developers to assign tasks to agents and expanded its Rovo search, chat and automation facility to integrated third-party agents using the open Model Context Protocol. Executives said the goal is to shift AI from chat-based assistance to embedded, accountable participants in enterprise workflows.

“A lot of our customers have created agents in Rovo,” said Sanchan Saxena, head of product for the teamwork collection at Atlassian.  “Now those agents can come to where the work is happening and be much more proactive. It’s a natural evolution of a chat interface to an agentic interface.”

Rather than competing directly with standalone copilots, Saxena framed the move as addressing governance needs. “The challenge is no longer in finding or convincing people to use agents,” he said. “People have got many agents. They need to figure out problems like tracking who’s working on what, auditability, privacy and risk management.”

Jira’s existing controls provide that structure through features like audit trails and role-based access controls. “If I don’t have access to the company’s financials, the agent doesn’t have access either,” Saxena said. “Just like every human action is documented in Jira, every agent action is also documented.”

Data integrity is ensured by allocating a private sandbox to each individual developer and agent. Agents have no permission to delete or modify production code. Changes become permanent only after human approval.

A centerpiece of Atlassian’s strategy is its Teamwork Graph, a map of relationships among people, work and knowledge across Jira, the Confluence collaboration platform and code repositories. “It is the institutional knowledge of a company,” Saxena said. That gives Atlassian agents an advantage against third-party copilots because they have more institutional knowledge.

The expanded Rovo MCP Gallery allows third-party agents from GitHub Inc., Box Inc., Figma Inc. and others to be integrated with Rovo. Tools in the gallery can pull live data, take actions like creating a task or updating a record and combine external context with Atlassian knowledge for better reporting.

“We are big believers in MCP,” Saxena said. Atlassian said nearly one-third of agentic MCP operations generated by its customers are writes, indicating that agents are being used for active collaboration rather than just data extraction. It said 93% of MCP usage comes from paid tiers and enterprise customers account for nearly half of MCP usage.

The Gallery is launching with a limited number of qualified third-party agents and will be expanded over time.

The broader objective, he said, is to prevent “agent sprawl” from turning into chaos. “You will have hundreds of thousands of agents,” he said. “We will turn that into teamwork so you can manage, govern and secure those agents clearly.”

Image: SiliconANGLE/Ideogram