SNACK three-line summary
- GitHub opened a Claude agent provider preview in its June 22 Copilot update for JetBrains. The direction is to expand the choice of agents without leaving the IDE.
- The same update also bundles organization and enterprise-wide shared agent deployment, message steering during Copilot CLI runs, general availability of Cloud agent, and AI credit display per turn.
- However, the Claude agent currently runs in bypass permissions mode, and Copilot Business and Enterprise users need administrator policy activation. This does not mean every developer’s default workflow has changed overnight.
Snackgirls editor note
AIKO: “For coding AI now, it feels less like ‘one model does everything’ and more like whether you can choose the right agent for the task and change direction midway.”
Red: “The big part is especially being able to step in with Add to Queue or Steer with Message when a CLI task runs long. It clearly creates a point where human hands can come back into the agent’s work.”
What changed
GitHub said it added a preview feature for choosing Claude as an agent provider in its June 22 Copilot update for JetBrains IDEs. The flow is to connect the Claude Code CLI path in settings, then select Claude from the agent picker in Copilot Chat.
The update also bundles several features at once: the ability to use shared custom agents defined by organization and enterprise administrators directly inside JetBrains IDEs, the ability to queue a new message or change direction during a Copilot CLI run, a session log summary screen, improvements to the model selector, and AI credit display per turn.
Why it stands out
This update shows that competition among AI coding tools is no longer only about “autocomplete one line of code.” GitHub is pushing both a flow where organizations deploy shared agents to align team standards and a flow where users can intervene midway in agent tasks that are already running.
Put simply, coding AI appears to be moving away from a window where you receive a single answer and toward a workboard for operating multiple agents and task rules. For JetBrains users, that change has now arrived directly inside the IDE.
Points for general readers and working developers
For general readers, it may be confusing why someone would choose Claude again inside Copilot. As an analogy, it is like gaining the ability to swap helpers within the same workshop. GitHub showed a direction of increasing flexibility by allowing users to choose a different agent style within the Copilot framework.
For working developers, administrator-deployed agents and runtime steering features matter more. They can reduce the inconvenience of having to simply wait through long CLI tasks, while also letting teams distribute approved agent configurations at the organization level.
What still needs caution
According to GitHub’s announcement, the Claude agent currently runs in bypass permissions mode. In other words, file edits and tool calls are automatically approved, so it should not be treated as a feature that can simply be turned on without concern.
Copilot Business and Enterprise users also need an administrator to enable the Editor preview features policy, and Claude Code CLI must be installed locally. So this news is less a full official switchover and more a public preview showing where agent-style coding tool operations are headed.
Sources and checked date · Announcement 2026-06-22 / Checked 2026-06-23T01:30:12+00:00
Sources