Summary: On October 10, 2026, GitHub shipped an update to Copilot for JetBrains that gives enterprise admins control over default models, adds a one-click Fix action in diagnostic menus, and lets developers disable automatic MCP server startup. There are also reliability fixes and a deprecation: JetBrains IDE 2025.1 is no longer supported. If you ship inside an IDE plugin or manage Copilot across a team, the interesting part is how much of this release is about governance knobs rather than new intelligence.
The admin gets the default model
The headline change is enterprise-managed default models. Administrators can use managed settings to pick any available Copilot agent model as the starting model for new conversations. Individual developers can still override it with the model picker, but the org decides the baseline.
That matters for two practical reasons. First, default settings quietly shape behavior — most people never change the picker, so the enterprise default effectively sets your team’s cost and capability profile. Second, it closes a governance gap: today an org can require a specific model, but the starting point for every new chat has been whatever the IDE last used. If you manage Copilot across a company, this is worth configuring deliberately rather than leaving to habit.
Fix actions and MCP startup control
Two smaller changes hit daily workflow. Diagnostic intention menus now include a Fix action that opens inline chat with Copilot asked to repair the reported issue — agent mode when available, ask mode otherwise. Fewer steps from lint error to proposed patch is the kind of friction reduction that adds up over a workday.
The MCP setting is the one I’d flag for anyone running lots of local tooling. You can now disable automatic server startup for Copilot and Claude, so configured tools only become active when you decide. Automatic startup is convenient until you have a dozen servers spawning on IDE launch, each consuming context and memory before you’ve asked for anything. Deferring startup is a reasonable default for most setups.
The unglamorous half: reliability and deprecation
The rest of the release is housekeeping, but it’s the kind that determines whether people keep the plugin installed. GitHub reports fixes across language server startup, model and provider switching, MCP configuration, customization refreshes, and worktree workflows. Chat got usability polish too — messages grouped by turn with turn-to-turn navigation, copy and edit actions on user messages, clearer account labels, and a Switch Account action. New users get a one-time reminder pointing them to Chat, and the sign-in page now supports cancellable browser authorization.
One deadline to note: the plugin requires JetBrains IDE 2025.2 or later, dropping support for 2025.1. If your team pins IDE versions, check before the update rolls out.
Why the governance knobs are the story
Zoom out and this release is mostly about who decides what, not what the model can do. Admin-set defaults, opt-in MCP startup, clearer account switching — these are controls that make Copilot something a company can adopt as policy rather than something individual developers install on their own.
That tracks with a broader pattern: the teams getting value from AI coding tools are the ones treating configuration as a product decision, not a personal preference. I wrote about a similar idea recently in why two-tier token pricing and effort knobs change how you route cheap work — defaults and routing decisions made upstream shape outcomes far more than per-prompt choices downstream.
The practical takeaway: if you admin a GitHub org, the managed default model setting is live and cheap to set — pick a baseline that matches your cost posture. If you’re just a JetBrains user, the Fix action and the MCP startup toggle are the two features worth adopting this week. And the supplied changelog doesn’t specify rollout timing for managed settings across plans, so confirm availability with your GitHub contact before promising it to a team.
Sources
View the original post on X