Atlassian Launches Agentic Multiplayer Protocol to Put AI Agents Alongside Human Workers

Quick Facts

• Atlassian's new Agentic Multiplayer Protocol (AMP) gives AI agents distinct identities, scoped permissions, and real-time presence alongside human teammates in Jira and Confluence.
• The company's Teamwork Graph now connects more than 250 billion objects and relationships, with agents grounded in it delivering 44% more accurate answers while using 48% fewer tokens.
• Humans and AI agents interact on the Atlassian platform more than 10 million times per month, with Rovo MCP Server reaching 1 million monthly active users.

Atlassian Corp. announced the Agentic Multiplayer Protocol on Wednesday at Team '26 Europe. The protocol sets the rules for how AI agents participate in shared work alongside humans on Atlassian's platform.

AMP defines how agents carry an identity, operate under scoped authority, access shared context, and produce results that humans can review and audit. Every agent, whether Atlassian's own Rovo or a third-party system, gets its own non-human identity, an owner, and a visible presence in real-time collaboration spaces.

Three Pillars of AMP

The protocol rests on three areas. First, agents can join work wherever it happens, responding to @mentions in Confluence pages, Jira comment threads, or Loom video briefs. The Atlassian MCP Server pulls in context from external tools including Figma and developer IDEs.

Second, agents appear in presence bars and cursors on canvases and whiteboards, the same way human collaborators do. A feature called Agent Sessions surfaces agent activity inside Jira and the Teamwork Graph so work context is not lost.

Third, AMP enforces strict governance. Agents can run under a user's own permissions or through dedicated service accounts, and every action generates an audit trail. Agents draw context from the Teamwork Graph, which now spans more than 250 billion objects and relationships.

Human Oversight Stays Central

Head of AI Product Jamil Valliani told SiliconANGLE the company holds an internal position: "Headless software means brainless software." Atlassian does not want agents operating without human direction.

"This is a multiplayer game," Valliani said. "Humans and agents are working together in a very dynamic space, and there are lots of players."

Rovo, Atlassian's AI assistant, can execute tasks over hours, but users set the goal, review the proposed plan, and approve the final output. Autonomy is built around checkpoints, not unchecked action.

New Products Ship Alongside AMP

Atlassian also announced Rovo Work, a new mode in Rovo Chat that handles complex, multi-step tasks inside a secure sandbox. Users set an objective, the system proposes a plan, and the user shapes it before execution begins across Jira, Confluence, and connected tools.

A new Artifacts app gives AI-generated content from any tool a permanent URL that embeds into Confluence pages and Jira tickets. The company shipped more than 120 new capabilities total, including an upgraded MCP Server that cuts token usage for Jira and Confluence tasks by 25%, a Code Search app, and expanded Teamwork Graph connectors for Zoom, Gong, Microsoft Entra ID, and Google Identity.

Scale and Enterprise Adoption

CEO and co-founder Mike Cannon-Brookes said context is the competitive edge companies cannot hire. "Thanks to 25 years of connecting teams, customers get one of the best context graphs for orchestrating agentic workflows," he said.

Rovo MCP Server monthly active users doubled in one quarter to reach 1 million. Rovo assisted actions grew 50% over the prior quarter. The company reports that 80% of Fortune 500 companies have adopted its AI platform.

Chief Product and AI Officer Tamar Yehoshua said the shift is direct: agents are now doing work that humans once did alone, and coordinating between the two is the core product challenge Atlassian is solving.

Read more: Atlassian lays groundwork for humans and AI agents to work side by side