GitHub Copilot for JetBrains: MCP server toggles and default-model controls land in the IDE

What it is

GitHub Copilot for JetBrains is the Copilot plugin that runs inside IntelliJ, PyCharm, GoLand, Rider and the rest of the JetBrains family. The October 10 update is small but relevant if you work on agent loops: it brings more control over default models and MCP server in GitHub Copilot for JetBrains, plus some chat and reliability fixes. Think "configuration knobs," not "new capability."

The next-work-session test

Concrete scenario: you're wiring a local SDK through an MCP server — say a database tool and a tracing tool — and you're A/B-ing two models on a replayed eval set. Before this update, every time you opened the IDE, Copilot's MCP servers spun up whether you wanted them or not, and your default model was whatever the picker last remembered.

Now: a new MCP setting lets you disable automatic server startup for Copilot and Claude, giving you more control over when configured tools become active. That means you can open the project, run your eval harness against a clean baseline, then flip MCP on when you actually want tool-calling in the loop. For eval work where tool availability is a variable you're trying to control, that's the useful bit.

The default-model piece is aimed at admins, not you: enterprise administrators can now use managed settings to choose any available Copilot agent model as the default for new conversations, which gives organizations more control over the starting model while still preserving an explicit model picker choice. If you're on a team plan, expect your "default" to be set for you — but you can still override per conversation.

Pricing

The plugin itself is free; you pay for a Copilot seat. Based on public pricing pages as of 2026, GitHub lists Copilot Free at $0, Pro at $10/month, Pro+ at $39/month, Business at $19/user/month and Enterprise at $39/user/month, with AI Credits included on paid plans. The enterprise-managed default model control in this update only applies if your org is on a plan where admins can push managed settings — effectively Business or Enterprise. Treat third-party pricing summaries as a starting point and confirm on your billing page.

What we'd actually use it for

Honestly, two things:

That's it. We wouldn't frame this as "better agent loop" — it's "fewer surprises in the agent loop."

Limits

• The default-model control that actually sticks across the org is admin-only. Individual devs still rely on the per-conversation picker.
• Support for JetBrains IDE 2025.1 has ended as of the October 10 Copilot update, so teams still on 2025.1 won't receive the new default model controls, Fix action, or MCP startup settings until they upgrade their IDE. If your team pins IDE versions, check this before planning around the new toggles.
• MCP management inside the JetBrains plugin is still catching up to the VS Code experience. The previous release already added some of this: Copilot for JetBrains 1.18.0 brings AI-assisted tool approvals, more control over agent conversations, and shared skills and instructions for your organization, so if you skipped 1.18, you're getting two updates' worth of controls at once.
• No new models, no new eval tooling, no observability improvements in this changelog beyond general reliability notes.

Try it if

• You run MCP servers locally and want them off by default at IDE launch.
• You're an admin trying to standardize which agent model your team starts conversations on.
• You're on a current JetBrains IDE (2025.2+) and already have a Copilot seat.
• You're mid-migration from the VS Code Copilot experience and want JetBrains parity on MCP controls.

Skip it if

• You're not using Copilot's agent mode or MCP at all — this update won't change your inline-completion day.
• You're locked to JetBrains 2025.1 and can't upgrade.
• You were hoping for new models, a better evals surface, or deeper tracing — none of that is in this release. Watch the Copilot changelog feed instead and come back when something bigger ships.