Introducing the AI context engine for your entire codebase

Atlassian has introduced Code Context as part of its Teamwork Graph, addressing a key limitation faced by coding agents: the lack of contextual understanding beyond individual workspaces. The feature integrates codebase comprehension with organizational signals such as architectural decisions, product strategy, and cross-team dependencies, enabling agents to operate with greater precision. By indexing code into the Teamwork Graph, Code Context allows agents to retrieve relevant source code and related work items from tools like Jira and Confluence, ensuring they start from a well-informed position. This reduces the need for developers to manually explain code behavior or dependencies, streamlining the development process.

Internal benchmarks cited by Atlassian indicate that agents using the Teamwork Graph with Code Context delivered 44% more accurate results while using 48% fewer tokens compared to those without it. The feature is accessible via the Teamwork Graph CLI and supports coding agents such as Cursor, Claude Code, and Codex, which can query across multiple repositories and retrieve context from connected sources. Code Context also integrates with Rovo Chat, allowing developers to ask natural-language questions about system functionality and receive answers grounded in both code and organizational knowledge.

Security and governance remain central to Code Context’s design. Admins must opt in before indexing begins, and access is scoped to each user’s or agent’s existing permissions in source code management (SCM) systems like Bitbucket and GitHub. The feature respects existing SCM permissions, ensuring that only authorized code is retrieved. Atlassian emphasizes that the goal is to provide agents with the right access—not unlimited access—while keeping developers accountable for reviewing changes and downstream impacts before implementation.

Code Context, previously referred to as “Code Intelligence,” is currently available in early access and is rolling out gradually to customers through an open beta. Organization admins can check availability and enable the feature under Rovo settings in Atlassian Administration. Support is provided for Bitbucket and GitHub repositories, with documentation available for security considerations and setup guidance. Atlassian positions this enhancement as part of its broader effort to improve AI-powered development by leveraging machine learning to identify relevant Jira work items and provide contextual support.