Surprisingly Useful AI Article Enhancements
Atlassian has expanded access to its Teamwork Graph, the context engine underpinning its AI capabilities, making it available to third-party AI tools and agents beyond Atlassian’s own product suite.
The updates, unveiled at the company’s Team ’26 user conference, include a new command-line interface for developers, an MCP server for knowledge workers and general availability of custom connectors that let enterprises pipe data from proprietary systems into the graph.
The Teamwork Graph now holds more than 150 billion objects and relationships, drawing from Jira updates, linked content, connected tools and other sources to map how work happens across an organisation.
The context problem for AI agents
Atlassian is framing the updates around a core argument: that AI agents fail not because the underlying models are inadequate, but because they lack the context needed to produce useful results.
The company contends that information is typically scattered across tools, siloed by department and stripped of the human context – such as why a decision was made, who owns it now or what went wrong previously – that agents need to operate effectively. Without that context, agents guess or hallucinate, and teams end up working from different versions of the truth.
The Teamwork Graph is designed to address this by connecting people, goals, code and content across Atlassian products and other connected SaaS applications. Atlassian claims that grounding AI responses in Teamwork Graph data delivered 44 per cent more accurate results while using 48 per cent fewer tokens in the company’s internal benchmarks.
A CLI for developers
The Teamwork Graph CLI, now in open beta, gives developers a way to query graph context directly from the terminal and pipe it into any AI tool, agent or workflow they are building.
Atlassian is positioning the CLI as a way to let coding agents such as Claude Code and Cursor access work data and relationships through a single interface rather than stitching together calls to individual product APIs. The CLI offers more than 300 commands and supports both read and write operations, meaning developers can update work items and create relationships as well as query them.
Use cases include browsing entities such as people, projects and issues across platforms, running parameterised lookups and embedding graph queries into CI/CD pipelines, incident runbooks or governance scripts.
Administrators retain control over scopes and permissions.
MCP server for knowledge workers
The Rovo MCP Server, also in open beta, extends Teamwork Graph access to any AI assistant that supports the Model Context Protocol (MCP). This includes tools such as Claude Cowork and ChatGPT.
Atlassian describes this as enabling AI assistants to both read from and act on organisational data, rather than simply extracting information. The company acknowledges that these MCP and CLI integrations effectively function as “headless apps” – software operating through APIs and AI rather than a traditional user interface.
Practical applications include incident response agents that can pull together related Jira issues, deployments and past remediation steps for faster triage; notification agents that identify current project owners rather than relying on static watcher lists; and retrieval tools that navigate relationships across Confluence and Jira to surface authoritative sources rather than just relevant pages.
Custom connectors now generally available
For enterprises whose critical data lives in proprietary or industry-specific systems, Atlassian has made Teamwork Graph Connectors generally available through Forge, its cloud development platform.
Customers and partners can now build their own connectors to bring data from internal tools, legacy systems and specialised platforms into the graph with existing access controls preserved. Once connected, that data automatically becomes available across Rovo agents and Atlassian Analytics without further configuration.
Related: Best Business Laptops for work & school
Related: Best Gaming Laptops
Related: Best Portable Laptop
Mercedes-Benz is an early adopter of this capability. The automotive manufacturer built custom Forge connectors for its specialised systems covering defect management, requirements traceability and release workflows.
Magnus Ostberg, Chief Software Officer at Mercedes-Benz, commented on the results.
“Rovo is providing our team members with the right data exactly when they need it,” Ostberg stated. “Having agents find, connect and act on information from across our Atlassian platform is a great enabler.”
Atlassian reports that Mercedes-Benz has seen a 90 per cent improvement in defect intake quality, 85 per cent faster duplicate detection and tenfold faster software delivery as a result of the integration.
In-product signals for graph intelligence
Atlassian is also introducing new visual indicators across Jira, Confluence, Loom and Rovo that show users where the Teamwork Graph is adding value within their workflows.
These include related work signals that surface connected issues, pages and runbooks across applications; people signals that highlight relevant owners and recent contributors inline; and decision trails that expose upstream and downstream dependencies during changes.
The intent is to ensure that both AI agents and human users are operating from the same data, reducing the time teams spend searching for context and the risk of decisions being made on incomplete information.
A new destination for graph visibility
Alongside the product updates, Atlassian has launched TeamworkGraph.com, a web destination where users can explore what their Teamwork Graph contains and identify opportunities to strengthen it.
The site lets users see who they collaborate with most frequently, visualise relationships between people, projects and tools without writing code, and receive recommendations on connecting additional data sources to improve AI accuracy.
The Teamwork Graph CLI, Rovo MCP Server and custom connectors are available now. Further details are available through Atlassian’s Team ’26 conference sessions, which are accessible on demand.
Last Updated on May 10, 2026 by Nick Ross
