Atlassian Code Context Open Beta: Availability and Setup Guide

Quick answer

Atlassian Code Context adds lexical and semantic search across connected source repositories to Rovo Chat and the Teamwork Graph (TWG) CLI. As of August 27, 2026, Atlassian’s current support page says it is open beta for GitHub users and rolling out to Bitbucket shortly. The experience may change before general availability.

An organization admin must find and enable Code Context in Atlassian Administration. Eligibility is not universal: the setup guide lists Atlassian Cloud Standard, Premium, and Enterprise and says Atlassian Government Cloud is not available.

Current availability matrix

Question	Current official answer	What to verify
Lifecycle	Open beta; experience may change before GA	Support, change tolerance, rollback, and procurement status
GitHub	Open beta for GitHub users	Eligible site, GitHub Cloud connector, repository scope, and permissions
Bitbucket	Rolling out shortly on the current support page	Whether the enable control is visible for your site today
Surfaces	Rovo Chat and TWG CLI	User access, client version, third-party agent path, and data flow
Admin role	Organization admin enables it	Named owner, terms review, site selection, and disable procedure
Cloud editions	Standard, Premium, Enterprise	Actual subscription and site eligibility
Government Cloud	Not available	Do not plan a pilot there based on commercial-cloud docs

Sources: what Code Context is and setup instructions, reviewed August 27, 2026.

What Code Context does

Lexical search finds exact terms such as function names or strings. Semantic search ranks code by conceptual similarity when the query does not match the repository’s wording. Rovo Chat and the TWG CLI can choose or combine those modes.

The index can help people and agents locate implementations, dependencies, configuration, and related code across repositories. It does not prove that the top result is complete, current, safe, or the correct place to edit. Atlassian documents no confidence signal for semantic results and warns that recently pushed code may not appear immediately.

How indexing works

Once enabled, Atlassian connects through the Jira GitHub Connector or Bitbucket integration, chunks files, creates vector embeddings, and stores code chunks and embeddings in an isolated Atlassian-managed environment. Repository events trigger incremental re-indexing.

The current setup guide says initial indexing can take up to 24 hours and provides no indicator showing which repositories are indexed. Build a readiness check around evidence you can observe rather than assuming the whole source estate is searchable.

Review the GitHub vs. Bitbucket setup guide for connector and indexing differences.

Permissions are necessary but not sufficient

Atlassian says Code Context validates query-time access against the source-code-management permission endpoint and does not grant access beyond existing SCM permissions. That is an important retrieval boundary, but it does not answer every governance question.

Before enabling, confirm:

1. Repository membership and branch access are correct in GitHub or Bitbucket.
2. Your organization has the rights to mirror, index, store, and embed the selected code.
3. Repositories have been scanned for committed secrets and findings have been rotated.
4. US storage and processing fit data-residency and sovereignty requirements.
5. Third-party agent providers are approved before TWG CLI sends them retrieved source.
6. Generated answers and proposed changes still receive human review and normal repository checks.

The security, permissions, and indexing guide covers these boundaries in detail.

Do not treat Atlassian’s benchmark as your result

Atlassian’s launch reports 44% more accurate results and 48% fewer tokens in internal benchmarks. Those are vendor-reported results, not an independent guarantee for your repositories, prompts, agent, model, permissions, or tasks.

Atlassian also publishes a local blinded benchmark that asks the same agent and model the same questions with siloed context and Teamwork Graph context, then hides which answer came from which arm during review. Use representative tasks and report answer preference, token use, and time separately. The benchmark documentation calls results directional rather than final.

See the Atlassian Teamwork Graph benchmark and run the Code Context readiness check before an opt-in decision.

Pilot sequence

1. Confirm lifecycle, edition, site, SCM, and the visible admin enable control.
2. Scan selected repositories for secrets and document indexing rights and residency approval.
3. Correct SCM permissions before indexing.
4. Enable one low-risk repository set and wait for the documented indexing window.
5. Test lexical exact matches, semantic concepts, stale branches, denied repositories, and recent pushes.
6. Compare Rovo Chat and TWG CLI data paths separately.
7. Run a blinded benchmark with representative questions.
8. Test disablement, deletion expectations, connector revocation, and user offboarding.

Frequently asked questions

Is Atlassian Code Context generally available?

No. The current documentation calls it open beta and says the experience may change before GA.

Does Code Context support GitHub and Bitbucket?

Atlassian documents GitHub Cloud and Bitbucket Cloud support, but the current rollout wording differs: GitHub users are in open beta while Bitbucket is rolling out shortly. Check your admin console.

Does it let agents edit code?

Code Context supplies retrieval context. Editing authority comes from the coding agent, local tools, repository permissions, and your approval workflow. Preserve branch protection, review, checks, and human merge authority.

Is semantic search a security scanner?

No. Atlassian explicitly says it is a discovery tool, not a replacement for dedicated security or compliance scanning.

Official sources

• Atlassian: introducing Code Context
• Atlassian Support: what is Code Context
• Atlassian Support: set up Code Context
• Atlassian Developer: Teamwork Graph benchmark

Source check: August 27, 2026. Recheck lifecycle, editions, rollout, SCM support, connector requirements, indexing, storage, limitations, and data terms before use.