GitHub Copilot Brings Claude Agent Sessions and Governed Agents to JetBrains IDEs
GitHub's June 22 Copilot update for JetBrains IDEs adds Claude as an agent provider preview, organization and enterprise agents, queued CLI steering, debug summaries, and per-turn AI credit visibility.
GitHub Copilot's June 22, 2026 JetBrains update makes agent work inside IntelliJ-family IDEs easier to govern and interrupt. GitHub says the release adds organization and enterprise agents, Claude as an agent provider in public preview, queued or steering messages for Copilot CLI sessions, agent debug summaries, and per-turn AI credit indicators. The JetBrains Marketplace stable plugin listing independently confirms the updated Copilot plugin channel and its AI credit indicator language.
Key takeaways
- Organization and enterprise agents can now appear directly in the JetBrains Copilot agent picker when admins publish them at GitHub org or enterprise level.
- Claude as an agent provider is in public preview for JetBrains IDEs, but it requires the Claude Code CLI path to be configured locally.
- Copilot CLI sessions now support queued follow-ups, steering messages, and stop-and-send behavior during a running request.
- Local, CLI, and Claude agent sessions show per-turn AI credit usage, useful for teams managing GitHub's usage-based billing.
- GitHub notes that Claude currently runs in bypass permissions mode, so Business and Enterprise admins should check preview policies before enabling it.
Practical LinkLoot angle
This update matters most for teams standardizing agent workflows across many developers. A useful rollout pattern is to publish a small set of approved agents at the GitHub organization level, enable the JetBrains preview for a pilot group, and require developers to record when they switch from Copilot's native agent to Claude. That keeps experimentation visible while admins compare cost, review quality, and permission posture.
| Workflow choice | Best use | Limitation | Source |
|---|---|---|---|
| GitHub org or enterprise agents | Reusable team workflows in JetBrains IDEs | Admin setup and eligible user access required | GitHub Changelog |
| Claude agent provider preview | Trying Claude Code from the JetBrains Copilot panel | Requires Claude Code CLI; bypass permissions mode noted by GitHub | GitHub Changelog |
| Queued or steering CLI messages | Redirecting long Copilot CLI sessions without losing the current run | Teams still need norms for when to steer versus stop | GitHub Changelog |
| Per-turn AI credits indicator | Budget monitoring during agent sessions | Indicator helps visibility, but billing policy still lives in GitHub plan settings | GitHub Changelog and JetBrains Marketplace |
For a broader agent tooling checklist, keep this next to LinkLoot's guide to AI agent tools.
What to verify before you act
Check the installed JetBrains Copilot plugin version and release channel before assuming every developer has the June 22 features. For Claude sessions, verify the Claude Code CLI path, the Copilot Business or Enterprise editor preview policy, and whether bypass permissions mode is acceptable for the repository class being edited. For cost control, compare the per-turn indicator against GitHub's billing dashboards for a few real tasks before using it as the only spend signal.
Source check
GitHub's changelog is the primary source for the feature list, availability notes, and Claude preview caveat. JetBrains Marketplace confirms the stable plugin listing and independently surfaces the updated Copilot plugin context, including the per-turn AI credit indicator wording.
GitHub added organization and enterprise agents, Claude as an agent provider preview, CLI message queueing and steering, debug summaries, model picker improvements, and per-turn AI credit indicators.