Copilot for JetBrains adds tool controls and raises its IDE minimum

A broken shop page can start with a small warning in your editor. Before asking an agent to fix it, you need to know which model and connected tools it will use. GitHub’s October 10, 2026 JetBrains update gives you more control at that starting point, alongside a quicker route from a diagnostic to a proposed repair. Original release.

First check your editor: JetBrains IDE 2025.2 or later is now required. Support for 2025.1 has ended. GitHub directs users to the latest Copilot plugin; this is a plugin update, not a new public AI model.

Decide what starts with the conversation

Enterprise administrators can set an available agent model as the default for new conversations. A deliberate choice in the model picker remains respected. Separately, a new setting lets you stop MCP servers starting automatically for Copilot and Claude. MCP connects an assistant to configured tools. Model defaults and MCP control.

For a solo creator, the practical question is which connections a particular repair needs. A layout warning might need code inspection without every external integration running. Treat startup behaviour and account permissions as separate checks: the release does not describe the setting as revoking access.

Turn one warning into a reviewable change

The diagnostic menu’s new Fix action opens inline chat. It uses agent mode when available, otherwise ask mode. Diagnostic behaviour.

Here is a proposed check, not a test we ran: take one warning on that shop page, request a repair, then inspect the changed lines and open the page at phone width. A clean diagnostic alone would not prove the buy button still works. Keep the change only after the page’s actual task succeeds.

That is the useful next step after updating: choose one issue with a visible outcome and review its proposed fix. For the separate question of where an agent may write or run commands, read our Copilot local sandbox coverage.