Atlassian opens Teamwork Graph to third-party AI tools
Sat, 9th May 2026 (Today)
Atlassian has opened its Teamwork Graph to third-party AI agents and tools, extending its data layer beyond its own products.
The graph contains more than 150 billion connections across people, work, goals, code and content from Atlassian software and connected applications. Atlassian is making that context available through new open beta tools that let external AI systems search, reason and take action, while remaining subject to existing permissions and administrative controls.
The announcement marks a shift in how Atlassian wants its AI products used. Rather than keeping its Rovo assistant in a closed environment, Atlassian is positioning the underlying graph as a shared source of organisational context for a wider range of AI clients, coding tools and workflow software.
Mike Cannon-Brookes, Chief Executive Officer and Co-founder, outlined that argument in remarks released with the launch. "In 2026, anyone can buy 'smarts' by the token," he said.
"The real moat is your institutional memory: every plan, document, and decision your teams have ever made. Rovo is the interface that turns intelligence and context into actual momentum for your business."
Open access
One new route into the Teamwork Graph is an MCP server in open beta. It will allow Model Context Protocol-compatible clients, including tools such as Figma and Replit, to access information in the graph through a standard interface.
A second route is a command-line interface, also in open beta, aimed at software developers. Coding agents including Anthropic's Claude Code and OpenAI's Codex will be able to draw on graph data directly within development environments, while administrators retain control over access scopes and permissions.
That approach reflects a broader contest among software companies over where business context will sit as AI agents become more common in office work and software development. By exposing its graph through standardised access points, Atlassian is betting that control of structured workplace context will matter more than the model itself.
Atlassian tied the launch to growing use of Rovo among large customers. Rovo is already used by more than 75% of Fortune 500 companies and 90% of its enterprise cloud customers.
Customers carried out more than 14 million Rovo-assisted actions in the past month alone, while agent-based automations have increased sevenfold over the past six months.
Product changes
Alongside the graph opening, Atlassian introduced a broader set of AI-related product updates across Jira, Confluence, developer tools and product management software. The changes are intended to move Rovo from an assistant that responds to prompts toward software agents that can be assigned work or complete multi-step processes.
A new Rovo Studio workspace is now generally available, giving users a central place to create agents, automations and custom applications using natural language. Governance features are built in, an important point for larger customers that need to control how AI tools interact with internal systems and data.
Atlassian also detailed a feature called Max mode for Rovo Chat, due soon. It is designed to take a complex request, break it into an action plan and carry out tasks across Atlassian products and connected software services.
In Jira, agents are now generally available. Teams can assign tasks to them, mention them in comments and place them inside workflows and automations, giving AI tools a more explicit operational role in project tracking and execution.
Confluence is also being updated with a beta feature called Remix with Rovo, which lets users turn blocks of content into formats such as charts, databases and infographics without leaving the page. A slides feature for Confluence is due in beta later, extending that approach to presentation materials while keeping the source content in one place.
For software teams, Atlassian's DX AI Experience is now generally available. The feature is intended to help engineering organisations track where AI is generating code, assess how agents are performing and measure return on investment from AI use in the software development lifecycle.
The company is also rolling out an early access Product Collection, which brings together Jira Product Discovery, the Feedback app and other elements, including a planned Pendo integration. The aim is to give product teams one connected environment to capture feedback, rank ideas and follow delivery work.
Another element is Dia, Atlassian's AI browser, which is ready for teams to try. It is designed to produce a daily briefing based on emails, messages, calendars, graph context and browser activity, and includes single sign-on, mobile device management support for Chromium and SOC 2 Type II attestation.
Taken together, the updates show Atlassian pushing further into the workplace AI software market by tying agents to the records, workflows and documents inside its collaboration products. The central claim behind that strategy is that AI systems will be more useful if they can access an organisation's own history of plans, discussions and decisions.