GitHub Copilot Rolls Out Agentic AI Features for JetBrains IDEs

GitHub pushed its Copilot AI assistant into new territory on March 11, moving core agentic capabilities from preview to general availability for JetBrains IDE users. The update marks a shift from simple code completion toward autonomous, multi-step task execution—a feature set that’s been in development since Agent Mode first hit JetBrains around July 2025.

Custom agents, sub-agents, and the plan agent are now fully available across all Copilot plans. These tools let developers create specialized AI assistants tailored to specific workflows, delegate isolated sub-tasks for more precise execution, and structure complex problems into step-by-step plans. Think of it as moving from a code autocomplete tool to something closer to an AI pair programmer that can actually think through problems.

Agent Hooks Enter Public Preview

The more interesting development for power users: agent hooks are now in public preview. These allow custom commands to run at critical points during agent sessions—before and after tool use, when errors occur, or when a user submits a prompt. Developers can define hook configurations in a hooks.json file within their repository’s .github/hooks/ folder.

For enterprise teams, this opens doors for automated policy enforcement and integration with external tooling without losing control over what the AI actually does. Copilot Business and Enterprise subscribers will need an administrator to enable the Editor preview features policy first.

MCP Auto-Approve and Instruction Files

GitHub also added auto-approve support for MCP at both server and tool levels. The practical benefit: fewer manual approval clicks interrupting workflow during agent sessions. Configuration lives in Settings under GitHub Copilot’s Chat section.

On the customization front, Copilot now supports AGENTS.md and CLAUDE.md instruction files. The system automatically discovers and loads these files during agent workflows, pulling from the current workspace or globally. A new /memory slash command provides quick access to manage these preferences.

Model Selection Goes Automatic

Auto model selection is now generally available for all Copilot plans in JetBrains IDEs. Rather than manually picking which AI model handles your request, Copilot chooses based on real-time availability and performance. A new thinking panel for extended-reasoning models like Codex shows the model’s reasoning process, with configurable Anthropic thinking budgets.

A context window usage indicator now appears in the chat panel—useful for understanding when you’re pushing the limits of what the model can hold in memory during longer conversations.

Microsoft’s stock sits at $404.88 as of March 11, up 0.63% on the day with a market cap of $3.01 trillion. The company continues expanding Copilot’s capabilities across platforms, with the CLI version reaching general availability just last week on March 3. Edit mode has been deprecated in the chat mode dropdown, signaling GitHub’s bet that agentic workflows represent the future of AI-assisted development.

Image source: Shutterstock