GitHub Copilot has introduced significant updates to its agentic capabilities, making it easier to tailor the AI-powered coding assistant to individual workflows and collaborate with specialized agents directly within JetBrains IDEs.
Custom agents, sub-agents, and plan agent are now generally available, allowing users to create custom instructions for Copilot’s behavior. The update also introduces agent hooks, which enable users to run custom commands at key points during agent sessions, automating workflows and enforcing policies while maintaining control over agent behavior.
In addition to agentic capabilities, GitHub Copilot has expanded support for agent instruction files, made auto model selection generally available, and enhanced the model and reasoning experience. New features include a thinking panel for extended-reasoning models, improved user experience, and quality improvements.
The updates aim to improve stability, reduce interruptions during development, and provide a more intuitive approval workflow. Users can now configure auto-approve for MCP at both the server and tool level, reducing manual approvals in chat and keeping workflows uninterrupted.
To try out the latest version of GitHub Copilot, users can visit Settings > GitHub Copilot > Chat > MCP Server and Tool Auto-approve Configuration in their JetBrains IDE. Share feedback through in-product options or the GitHub Copilot for JetBrains IDEs issues repository to help refine and improve the product.
Source: https://github.blog/changelog/2026-03-11-major-agentic-capabilities-improvements-in-github-copilot-for-jetbrains-ides