GitHub Copilot for JetBrains adds agent and MCP controls
GitHub adds Copilot controls in JetBrains for MCP startup, agent model defaults and diagnostic fixes. Scope, support and limitations explained.
GitHub's October 10 update to Copilot for JetBrains adds managed default agent models, a switch for automatic MCP-server startup, and a Fix action in diagnostics. These are JetBrains plugin capabilities, not a blanket change to every Copilot product.
On October 10, 2026, GitHub published an update to Copilot for JetBrains IDEs focused on agent-session controls, MCP-tool startup and the path from a diagnostic finding to a suggested fix. These practical controls matter to organizations evaluating coding agents because autonomy, credentials and review boundaries are product- and environment-specific. The announcement describes the JetBrains plugin, not a universal change across every GitHub Copilot interface.
What changed exactly?
Enterprise administrators can now set any available Copilot agent model as the managed default for new conversations. GitHub says users retain an explicit model picker. This helps organizations standardize an initial choice without removing local selection, but it does not guarantee uniform model behavior across repositories, sessions or other Copilot clients. Teams still need to inspect their applied administrative settings.
Diagnostic intention menus now offer Fix. Selecting it opens inline chat with a request to repair the problem. GitHub says agent mode is used where available and ask mode otherwise. This makes the handoff from diagnosis to a proposed patch shorter, but it is not a correctness or security guarantee. Review the diff, run tests and approve changes before relying on an agent-generated fix.
Who benefits and what remains uncertain?
The release also adds a setting to disable automatic startup of configured MCP servers for Copilot and Claude. This is useful for deciding when connected tooling should become active. It should not be confused with a universal permission sandbox: individual MCP server access, credentials, network reachability and tool scopes remain separate controls. Organizations should verify each connection explicitly.
Additional changes improve account switching, chat navigation, agent-session reliability, model/provider changes and worktrees. The release also ends support for JetBrains IDE 2025.1: GitHub says version 2025.2 or later is now required. Developers using older versions should check compatibility before upgrading their agent tooling.
AgentenCode treats these details as scoped features of the JetBrains integration. Our related GitHub Copilot Coding Agent profile is useful background, but it does not prove that every JetBrains-specific control works identically in cloud agent sessions. The release description comes from GitHub's own changelog; AgentenCode has not independently measured productivity or security improvements.
Frequently asked questions
What does the MCP startup switch change?
GitHub says it can stop configured MCP servers from starting automatically for Copilot and Claude. Actual tool permissions need separate checks.
Which JetBrains IDE versions remain supported?
GitHub says Copilot support for 2025.1 has ended and 2025.2 or later is required. Check the current plugin compatibility listing.
Source verification
Primary sources and product scope
Official documentation. No independent benchmark or implied cross-product functionality.