This update brings several new features and improvements to GitHub Copilot in JetBrains IDEs. Core agentic capabilities, including custom agents, sub-agents, and plan agent, are now generally available, with agent hooks in preview and auto-approve support for MCP.

It also expands support for agent instruction files, makes auto model selection generally available, enhances the model and reasoning experience, and includes additional user experience and quality improvements.

New features

Agentic capabilities

Custom agents, sub-agents, and plan agent now generally available

Custom agents, sub-agents, and plan agent are now generally available in GitHub Copilot for JetBrains IDEs. These capabilities make it easier to tailor Copilot to your workflows and collaborate with specialized agents directly within your IDE. Learn more about creating a custom agent in JetBrains IDEs.

Agent hooks support in public preview

Agent hooks are available in public preview. Agent hooks let you run custom commands at key points during agent sessions, making it easier to automate workflows, enforce policies, and integrate with external tools while maintaining control over agent behavior.

You can define your hooks.json file with the name of your choice in the .github/hooks/ folder of your repository. Supported events include userPromptSubmitted, preToolUse, postToolUse, and errorOccurred hooks. Check out our documentation on agent hooks for more details.

Use agent hooks in JetBrains

If you are a Copilot Business or Copilot Enterprise subscriber, an administrator will have to enable the Editor preview features policy before you can use it.

Auto-approve support for MCP

You can configure auto-approve for MCP at both the server and tool level, reducing manual approvals in chat and keeping your workflow uninterrupted during agent sessions. To configure this, go to Settings > GitHub Copilot > Chat > MCP Server and Tool Auto-approve Configuration in your JetBrains IDE.

Auto-approve experience with MCP

Customization and instructions

Agent instruction file support

We have added support for AGENTS.md and CLAUDE.md instruction files, including the ability to generate an initial AGENTS.md file. This makes it easier to tailor Copilot’s behavior to your project’s conventions and workflows.

Copilot automatically discovers and loads agent instruction files that you’ve defined into context during agent workflows, from the current workspace or globally across all workspaces. This includes Copilot custom instruction files, file-based instructions, AGENTS.md, and CLAUDE.md.

Go to Settings > GitHub Copilot > Customizations to view and manage your instruction files, where you can also enable support for nested AGENTS.md and CLAUDE.md files. You can also create an initial AGENTS.md using the Generate Agent Instructions action in a new chat session.

set up your agent instructions files

/memory slash command

You can use the /memory slash command to quickly open settings for agent instruction files, giving you faster access to manage your preferences.

Model selection and reasoning visibility

Auto model selection now generally available

GitHub Copilot auto model selection is now generally available in JetBrains IDEs for all Copilot plans. With auto, Copilot can choose a model on your behalf based on real-time model availability and performance without manual switching.

Thinking panel for extended-reasoning models

A dedicated thinking panel is now available for extended-reasoning models, such as Codex, with configurable Anthropic thinking budgets. This gives you visibility into the model’s reasoning process.

Context window usage indicator

A new context window usage indicator in the chat panel lets you see how much context is being used during a conversation.

User experience

We have improved the sign-in experience to reduce friction and make Copilot more immediately accessible. The chat panel opens automatically after signing in for faster access to Copilot. And GitHub Enterprise users can benefit from a smoother login flow with device code prioritization.

Additional UX and workflow improvements include:

  • More responsive chat panel layout and cleaner context/file attachment handling.
  • Cleaner auto-approve UI for a more intuitive approval workflow.
  • Refined NES trigger timing and dismissal behavior.
  • Improved support for Windows ARM platforms.

Quality improvements

We’ve made a set of quality improvements to enhance stability and reduce interruptions during day-to-day development:

  • Improved stability when reading terminal output.
  • Resolved issues where file updates were not applied correctly by the replace_string_in_file tool.
  • Addressed UI inconsistencies, including lingering action buttons, blank chat panels, and MCP code vision appearing when signed out.
  • Reduced visual glitches such as file icon flicker when changing the selected code range.

Deprecation

Edit mode has been marked as deprecated in the chat mode dropdown.

Try it out

We encourage you to try out the latest version of the GitHub Copilot plugin and share your feedback. Your input is invaluable in helping us refine and improve the product.

Share your feedback

Your feedback drives improvements. We’d love to hear about your experience in the following channels: