Summary: GitHub shipped an update to Copilot for JetBrains on October 10, 2026, adding enterprise-managed default models, a setting to disable automatic MCP server startup, and a Fix action inside diagnostic menus. The plugin also got a batch of chat and account UX polish — and dropped support for JetBrains IDE 2025.1. If you manage Copilot for a team or build against MCP tooling, a few of these changes are worth acting on today.
Admins can now set the starting model
The headline change for teams: enterprise administrators can use managed settings to pick any available Copilot agent model as the default for new conversations. Individual users can still override it with the model picker, which is the right design — the org sets the floor, the developer keeps the choice.
The practical win is cost and consistency control. Instead of new hires landing on whatever model they picked once and forgot, the org chooses the starting point. It pairs naturally with the tiered pricing and effort-knob pattern other vendors have been shipping; I wrote about the same idea in Claude’s two-tier token pricing and effort controls.
MCP startup becomes opt-out
A new setting lets you disable automatic server startup for Copilot and Claude. Before this, configured MCP servers came alive whenever the plugin did — which is convenient until you’re debugging startup order, watching language server contention, or you simply don’t want certain tools running in a given workspace.
For anyone wiring up tool servers, this matters more than it sounds. Startup behavior is where most MCP configuration pain lives, and manual activation gives you a cleaner failure surface. The release also claims reliability improvements around MCP configuration specifically.
From diagnostic to proposed fix in one click
Diagnostic intention menus now include a Fix action that opens inline chat and asks Copilot to repair the reported problem. It uses agent mode when available and falls back to ask mode otherwise. That’s a shortening of a loop you probably run dozens of times a day: see the warning, copy the message, paste it into chat, add context, ask for a fix. Now it’s one menu item.
Smaller quality-of-life changes
The rest of the update is polish, but some of it touches daily friction:
- Signed-in accounts get clearer labels, a status display, and a Switch Account action — useful if you juggle personal and org accounts.
- Chat messages are grouped by turn, with controls to jump between turns, and user messages gained copy and edit actions.
- New users get a one-time reminder pointing at Chat, and the sign-in page was rebuilt with a responsive layout and cancellable browser authorization.
One hard cutoff to note: JetBrains IDE 2025.1 is no longer supported, so the plugin now requires 2025.2 or later. Check your team’s IDE versions before upgrading.
What to do with this
If you run Copilot across an org, the two highest-leverage moves are setting the managed default model and deciding your MCP startup policy before developers each solve it differently. The quality improvements — language server startup, model switching, worktree workflows — are the kind of thing you benefit from without configuring anything. GitHub is collecting feedback through in-product options and the Copilot for JetBrains issues repository, so flag rough edges while the release is fresh. And if you’re evaluating which coding assistant to standardize on, the admin-side controls here are increasingly the deciding factor, not the raw model quality — a pattern that shows up in hiring and tooling data alike, as covered in AI code growth and developer jobs.
Sources
View the original post on X