GitHub Copilot for JetBrains: Claude as Agent Provider Now in Public Preview, Plus Org Agents and CLI Message Queueing
Chris Harper
2 min read
Jun 22, 2026 · 21:11 UTC
TL;DR: The June 22 JetBrains release makes Claude selectable as an agent provider inside Copilot Chat — bypass-permissions by default — plus org-published agents, CLI message queueing, and per-turn credit tracking.
Four updates landed in the GitHub Copilot for JetBrains changelog on June 22:
Claude as agent provider (public preview). Select Claude from the Copilot Chat agent picker to start an agent session in your JetBrains IDE. Requirements: Claude Code CLI installed, its path configured in Settings. Current behavior: bypass-permissions mode — file edits and tool calls are automatically approved without a confirmation step. Granular permission controls are on the roadmap. Copilot Business and Enterprise accounts need admin approval to enable.
Organization and enterprise agents. Admins publish custom agents at the org or enterprise level; they appear in every member's Copilot Chat agent picker automatically — a clean way to standardize reusable workflows (style-enforced code generation, security review, onboarding scripts) across the team.
CLI message management. While a request is in-flight, the Send button becomes a dropdown: queue a follow-up message (runs after the current request), steer the active request with new input, or stop and start fresh immediately.
Model picker + per-turn credits. A new /models command gives direct access to the model picker, including larger context window selection. Local, CLI, and Claude agent sessions now show a per-turn AI credits indicator per interaction.
Why it matters: Claude Code inside JetBrains collapses the context-switch between IDE and terminal — but bypass-permissions mode is the default. Understand what the agent can auto-approve before enabling it on a production codebase.