Atlassian Bets That Context-Rich Agents Beat Context-Blind Ones

When Atlassian introduced Rovo AI in 2024, it wasn’t running solely on a foundation model. It was drawing on the Teamwork Graph, the company’s structured map of how people, teams, projects, and decisions connect across an organization, built over more than two decades of enterprise use.

That context layer is what Atlassian said set Rovo apart. Today, the company is extending that advantage outward. Announced at this year’s Team 26 conference, Atlassian is opening the Teamwork Graph to third-party AI agents and tools, a move that lets any MCP-compatible agent or automation reason over the same 150 billion connections that power Rovo itself.

In addition, Atlassian is releasing updates to Rovo Search, Chat, and Studio that push the suite from assistive AI to autonomous execution. Taken together, Jamil Valliani, the company’s vice president and AI product chief, tells The AI Economy that it’s all aimed at demonstrating how Atlassian’s context layer, combined with its AI intelligence, can deliver “massive acceleration.”

Opening up the Teamwork Graph

“The story of the Teamwork Graph is important for us to tell, first because we want to get that power to all of our customers, even when they don’t use Rovo,” Valliani says. “We want to make sure that that context is available for them genuinely in every use case they have. But it also highlights, in a key way, the things that only Rovo could do for you…our ability to go and actually endlessly access that context.”

The Teamwork Graph is not a simple index. It spans more than 150 billion objects and relationships—not just conversations and people, but active projects, open issues, Wiki pages, and external assets Atlassian ingests from across the enterprise: design files, code repositories, Google Drive, SharePoint, and more. Every time a customer links to an outside tool, Atlassian pulls that object into the graph, building a continuously expanding map of how work actually happens, and one that agents can now query directly.

“The context is really what is going to differentiate the apps that can provide the best AI quality from those that are merely kind of [using] a foundation model, which are becoming increasingly commoditized now,” Valliani says. He notes that Atlassian tools using the Teamwork Graph have seen a 44 percent improvement in answer quality and are using half as many tokens “because they’re able to actually operate a lot more efficiently and get what they need rather than just a bunch of random data that pollutes the context window.”

“We want to get that power into everyone’s hands,” he adds.

There are two ways Atlassian is doing this: The first involves an MCP server that’s currently in open beta. The second is designed for developers, admins, and coding agents who prefer working closer to the metal—a command-line interface. This is a single, structured entry point for working across code, incidents, documents, and goals with the full weight of the Teamwork Graph behind it.

Atlassian is also integrating the Teamwork Graph and Rovo directly into Microsoft Teams and Copilot via MCP. Users working inside Microsoft’s ecosystem can now access Atlassian’s context layer without leaving the team communication app or AI chatbot. It’s a practical acknowledgment that enterprise teams don’t live in a single platform, and the Teamwork Graph is only as useful as the places it can reach.

“We really view it as our best job to provide any agent, any tool that customers want to use, [with] access to this graph,” Valliani says. “It’s customers’ data, and we want to make sure that they’re able to benefit from all of the investments they’ve made in Atlassian and our platform as well, to make sure that they build the best apps possible that meet their business needs.”

Atlassian’s approach stands in contrast to how Salesforce handled opening up Slack’s data graph—a process marked by developer friction and fits and starts before MCP access arrived. Valliani was direct about the difference: for Atlassian, opening the platform was never a controversial decision—it’s part of its DNA.

He emphasizes that the Teamwork Graph wasn’t built in response to the AI moment. It predates it by years. “We fundamentally believed years ago that this is an important asset,” Valliani says. “We built it in a rigorous way—from day one, from the ground up—to meet all those enterprise-grade needs.”

When Slack restricted developers from indexing, copying, or permanently storing messages through its API, the company framed it as a security measure. But the move was widely read as a deliberate squeeze on Glean and other enterprise search apps that had built businesses on top of Slack’s data.

Valliani sees it differently. Rebuilding something like Jira or Slack from scratch, he said, would be prohibitively expensive—developers aren’t going to try. What they want is to extend what already exists. “We want to make it as natural as possible for them to leverage that data,” he says, “and transform it to do their own extra things.”

For Valliani, the message to customers is straightforward. First, look at every workflow your organization currently manages through Atlassian and find ways to accelerate it. Second—and perhaps more telling—even where Atlassian tools aren’t in the picture, the Teamwork Graph still is. “We want them to think about all the new opportunities they have to leverage that context to solve new problems or rethink their business processes,” he notes, “knowing that they can use that context with whatever AI tools they’re using in those different departments or areas as well.” After two decades of building, Atlassian is finally opening the door.

Changes Coming to Rovo

Alongside the Teamwork Graph announcement, Atlassian is rolling out updates to Rovo, its enterprise AI suite built around three core components: Search, Chat, and agents. Together, they’re designed to help organizations surface knowledge, automate workflows, and improve decision-making. The company says customers performed over 14 million Rovo-assisted actions last month alone, though Atlassian did not provide a baseline for comparison.

Starting with Rovo Search, Valliani says it can search Jira more effectively—he estimates it’s up to 40 percent faster. And when you’re in Jira, it can search across the broader ecosystem through a single interface, pulling in Figma files, Google Docs, and other connected tools.

As for Rovo Chat—the conversational bot on Atlassian’s platform—it’s receiving a new usage mode called Max. It can already be used in quick mode for fast, snappy answers that don’t require much context, and in a deeper thinking mode, which uses state-of-the-art models to generate detailed answers over time. Atlassian created Max mode to tackle complex multi-step workflow requests, such as writing code or performing specific data analysis.

“Max mode means exactly what it sounds like,” Valliani explains. “We’re going to throw everything we have at it. That means Rovo will even spin up a virtual machine in the cloud, write Python code if needed, and do anything required to take whatever task you give it, break it down into steps, work back and forth with your team to make sure the plan is good, and then it will learn how to do what you need it to do, even if it doesn’t have the tools in-house.”

To illustrate, Valliani described asking Rovo Chat to create a podcast briefing from a set of Confluence pages. With no prior instructions on how to produce a podcast, the app learned the task on its own, writing the necessary code, researching best practices for format and length, and ultimately producing an audio file. “That’s just a small taste of the magic that’s possible,” he says.

Max mode isn’t only available for Rovo Chat—it’s also available to third-party agents that integrate with the service.

Lastly, Atlassian is making Rovo Studio generally available. First announced in April 2025, Rovo Studio is a no-code/low-code platform for developing custom AI solutions, including AI agents, automation workflows, real-world objects with assets and schemas, and interactive content views.

But Rovo Studios’ status change isn’t the only thing that’s new. Atlassian is updating it so that all it takes to get started is to describe what you want built using natural language prompts.

Valliani reveals that since its introduction, there has been more than a 7x growth in the number of automated workflows or agentified using Studio-built agents.

Other Product Updates

Finally, here’s a rundown of all the other news being made at Atlassian’s Team 26 conference:

• New Product Collection: An expansion of Jira Product Discovery, a tool to capture and prioritize customer signals, with two additions: Feedback—an AI-powered intake layer that captures and synthesizes customer signals directly into prioritization (early access); and JPD Enterprise—provides portfolio-level governance for organizations managing oversight across multiple product lines (now generally available).
• Teamwork Collection Updates: Agents in Jira is now generally available, enabling agents to tackle real work items with full audit trails and administrative controls. The new Remix with Rovo and Confluence slides feature can turn any page into charts, timelines, org charts, or full presentation decks. Agent Briefings in Loom enable multimodal walkthroughs converted into structured prompts and suggested action plans, all with one-click Jira creation.
• Service Collection Updates: Atlassian is adding an Incident Command Center, a unified hub for detection, investigation, mitigation, and resolution, supported by Rovo-powered root cause analysis (RCA) across observability and deployment data. Also, there’s Rovo Service, an autonomous or supervised agent that provides L1 support.
• Developer Experience Updates: Three new additions, starting with Agent Experience, which provides visibility into what agents encounter—requirements clarity, codebase predictability—to improve success rates; AI Code Insights gives organizations a system-wide view of AI-generated code down to the commit level; and AI Pulse delivers proactive productivity signals to frontline engineering managers.
• New Dia Reports: Born out of Atlassian’s acquisition of The Browser Company, this delivers proactive, browser-native briefings—personalized reports ranging from interview prep to decision briefings—by combining the Teamwork Graph context with everyday tools users are already working in. The aim: Show users what they need before they think to ask for it.

The announcements come less than a week after Atlassian posted its third-quarter earnings, in which CEO Mike Cannon-Brookes said the company’s AI-powered platform was helping customers sign “bigger, longer-term commitments.” While today’s releases span the breadth of Atlassian’s platform, the centerpiece is the opening of the Teamwork Graph to any agent, on any platform. It’s Atlassian’s bid to ensure that its growth momentum becomes harder to reverse—with two decades of enterprise context as its moat, the company is betting that’s reason enough for customers, old and new, to stay.

Featured Image: Credit: Microsoft Copilot