Your AI coding agent is running blind. It sees the files you open, the repo on your machine, the snippets you paste in — and nothing else. For a two-file script, that’s fine. For an enterprise with forty repositories, interdependent services, and years of architectural decisions buried in Jira tickets and Confluence docs, that context gap produces broken code and wrong assumptions. Atlassian’s Code Context, announced August 12, is a direct answer to that problem.
What Code Context Does #
Code Context is a new capability inside the Atlassian Teamwork Graph that indexes your connected codebases — GitHub and Bitbucket repositories — and makes them queryable by developers and AI agents alike. Once enabled, agents can run exact searches, natural-language queries, and semantic retrievals across every repo in your org, all in a single call.
That’s the headline feature. The less-obvious one is what gets included alongside the code: Jira work items, Confluence pages, Loom recordings, and 50-plus third-party integrations. An agent asking about your authentication service doesn’t just see the code — it can find the Jira ticket that explains why the auth flow was redesigned in Q3, the Confluence page documenting the original architecture decision, and the related PRs that shipped the change. That’s the difference between a tool that generates code and one that understands it.
The Numbers #
Atlassian published internal benchmark results alongside the announcement. Agents enriched by the Teamwork Graph delivered 44% more accurate results while using 48% fewer tokens compared to agents operating without that context. The token reduction is counterintuitive at first — more context should mean more tokens, not fewer. The explanation: when agents can retrieve precisely relevant information via semantic search, they stop burning tokens on broad, speculative context dumps. Focused retrieval beats carpet-bombing the prompt.
How to Set It Up #
Code Context surfaces through the Teamwork Graph CLI. Installation is a one-liner:
bash <(curl -fsSL https://teamwork-graph.atlassian.com/cli/install)
After install, authenticate with your Atlassian account and add agent skills — context files that teach your coding agent how to query Atlassian data. Skills write to ~/.agents/skills
, a universal directory read by Cursor, Codex, Gemini CLI, GitHub Copilot, Claude Code, and Rovo Dev. Set it up once, and every agent you use picks it up automatically.
Organization admins enable Code Context in Atlassian Administration under Rovo settings. Repository access stays scoped to existing SCM permissions — agents can only retrieve code that the user is already authorized to see. No new permission surface, no new security exposure to configure from scratch.
Why the Permissions Model Matters #
Atlassian’s Rovo has faced security scrutiny this year — a prompt injection vulnerability allowed exfiltration of Jira and Confluence data via crafted links, and a content-based variant remained unresolved as of early August. Code Context is a different product surface with a different trust model: it’s admin-controlled, permission-scoped at the SCM level, and doesn’t introduce new retrieval pathways beyond what users could already access. Worth knowing, especially if your team has been cautious about Rovo adoption.
The Bigger Picture #
GitHub Copilot Workspace offers cross-repo planning. Sourcegraph Cody uses Sourcegraph search for large-repo context. The TWG CLI is open source on GitHub. None of them bring Jira, Confluence, and org-level work history into the same query. That’s Atlassian’s moat: the Teamwork Graph holds 150 billion objects and relationships built over a decade of software teams actually doing work. That’s not something competitors reconstruct quickly.
The honest take: Code Context is the most practical solution to the agent context problem for teams already running on Atlassian. It doesn’t require migrating your repos, switching your agent, or restructuring how your team works. It adds context to the agents you already use, through a tool stack your org already has. If you’re on Jira and GitHub or Bitbucket, this is worth enabling this week.