GitHub published a June 22, 2026 Copilot update for JetBrains IDEs that matters for QA engineers using IntelliJ IDEA, PyCharm, WebStorm, or other JetBrains tools for automation work. The headline change is that Claude as an agent provider is now in public preview inside GitHub Copilot for JetBrains IDEs.

The release also adds support for GitHub organization and enterprise agents in JetBrains IDEs, lets users queue or steer messages while a Copilot CLI request is still running, adds an agent debug logs summary view, improves model selection, shows per-turn AI credit usage, and makes Copilot cloud agent generally available.

What GitHub announced

  • Claude agent provider preview: JetBrains users can configure the Claude Code CLI path and select Claude from the Copilot Chat agent picker.
  • Organization and enterprise agents: admins can publish curated agents that are available to eligible users in JetBrains IDEs.
  • CLI steering: Copilot CLI sessions can queue a follow-up, steer the current request, or stop and send a new message.
  • Agent debug summary: the Agent Debug panel now gives a consolidated summary of session activity.
  • Per-turn credits: local, CLI, and Claude agent sessions show how many AI credits a turn consumes.
  • Cloud agent GA: Copilot cloud agent is no longer behind the Editor Preview feature flag.

Important preview caveat

GitHub’s note is important for teams with strict repository controls: the Claude agent currently runs in bypass permissions mode, meaning file edits and tool calls are automatically approved. GitHub says configurable permissions are planned for a future release. Business and Enterprise users also need an admin to enable the Editor preview features policy before using the preview.

Why this matters for QA engineers

For QA engineers, this update is less about trying another chatbot and more about bringing agent workflows closer to daily automation work. Many SDET teams live in JetBrains IDEs while maintaining Java, Kotlin, Python, JavaScript, or TypeScript test suites. Having governed agents and provider choice inside the IDE can make AI-assisted test maintenance easier to standardize.

  • Reusable QA agents become practical: an organization can publish agents for Playwright review, Selenium locator cleanup, API contract checks, or flaky-test triage.
  • Long-running CLI tasks are easier to supervise: queue and steering controls help when an agent is reviewing a large test failure or refactoring fixtures across modules.
  • Credit visibility supports governance: per-turn AI credit indicators help QA leads understand which agent workflows are expensive before scaling them.
  • Debug summaries improve trust: agent activity summaries make it easier to review what an agent actually did during a session.

A practical rollout check for QA teams

If your team wants to test the GitHub Copilot JetBrains Claude agent preview, start with low-risk automation repositories or branches. Avoid giving the preview direct access to production release branches until your team has reviewed the bypass-permissions behavior and repository policy settings.

  1. Create one narrowly scoped QA agent, such as “review Playwright test diffs for weak assertions.”
  2. Run it on a sample pull request and compare its suggestions with a human review checklist.
  3. Inspect the agent debug summary and any changed files before merging.
  4. Track AI credits per task type, not just total team usage.
  5. Document when testers should use Claude, Copilot local agent, or cloud agent for different workflows.

Bottom line

GitHub’s June 22 update expands Copilot for JetBrains from chat assistance toward more governed, inspectable agent workflows. The QA opportunity is clear: standardize repeatable review and maintenance tasks. The QA risk is also clear: preview agents with automatic approvals need tight boundaries, review discipline, and credit monitoring.

Sources