Atlassian and OpenAI expand partnership around Rovo agents and enterprise context

Atlassian and OpenAI are expanding their partnership to bring OpenAI frontier models across Atlassian's platform and Rovo. The agreement connects model capabilities with Atlassian's Teamwork Graph, an enterprise context layer linking people, projects, documents and decisions.

In OpenAI's October 6 announcement, the companies describe agents that use this context to help teams plan and deliver work. The expanded access names GPT-6 Astra and the GPT-5.6 series. The announcement does not provide a product-by-product rollout timetable or say that every customer has already received every integration.

The practical focus is bringing project information into the tools where a team already works. A model's answer about launch readiness can be more useful when it has the relevant tickets and decisions, but access to that information still depends on the permitted connections and the state of the records.

Rovo combines models with project information

OpenAI describes a product manager asking Rovo whether a launch is on track. In the example, Rovo connects Jira tickets, Confluence documents and discussions to identify engineering blockers, missed milestones and decisions requiring attention. The models then help prepare a readiness assessment and recommended next steps.

That is an illustrated workflow, not a published evaluation of how often Rovo catches a real launch blocker. Teams adopting it should check whether the assessment cites the relevant work items and distinguishes documented decisions from inferences. An incomplete ticket or stale document can remain a weak input even when a model produces a polished summary.

Through OpenAI APIs, Atlassian can incorporate newer reasoning capabilities into Rovo as models advance. The companies present the agreement as ongoing access to changing model capabilities rather than a one-time connection to one fixed model. That makes permission handling and workflow review ongoing responsibilities too.

Codex adoption and plugins extend the connection

The collaboration began in 2023, according to OpenAI. Atlassian has since broadened its use of Codex and ChatGPT Enterprise. OpenAI reports that more than 3,000 Atlassian developers use Codex across terminals, integrated development environments and code-review workflows.

That usage count establishes reported adoption inside Atlassian. It does not measure a productivity improvement or prove that a particular code-review process has become safer. Those outcomes require separate evidence about the tasks and results.

The announcement says Atlassian plugins powered by Teamwork Graph let Codex users access work items and technical documentation. Atlassian and Teamwork Graph CLI plugins also connect ChatGPT and Codex to project information, documentation and development context, subject to appropriate permissions.

Atlassian's recently launched plugin extension brings Jira items, Confluence content and people into prompts. Its pinned Atlassian Home also surfaces assigned work, recent Looms, projects and Bitbucket pull requests. The account describes several routes for supplying context; it does not make them interchangeable with unrestricted access to everything in a company's workspace.

Deeper Jira integrations remain under exploration

OpenAI says the companies are exploring integrations that would let teams assign work to agents, track progress, capture decisions and review results through Jira. It pairs that direction with DX, Atlassian's platform for measuring developer productivity and engineering performance.

The proposed combination could help leaders assess development speed, cycle time and developer experience. The announcement does not provide measured improvements on those indicators or a general availability date for the exploratory features.

OpenAI also says it will continue using Jira for critical workflows within its own company. For customers, the near-term evaluation should focus on which connections are available, which project records an agent can retrieve and how a human reviews its proposed action. The partnership supplies a direction and access agreement, while workload-specific benefits still need to be demonstrated.