GitHub Copilot for JetBrains Adds a Default Model Setting and More MCP Control

GitHub has updated its Copilot plugin for JetBrains IDEs. The release adds new settings for the default model and the MCP server, makes diagnostic errors and warnings easier to work through, and improves chat navigation and account management. The announcement appeared in the GitHub Changelog on 10 October 2026. The text is short, and its list of changes breaks off at an ellipsis, so what follows covers only what the post names explicitly, plus a few cautious inferences.

Default model and MCP: more levers

Copilot stopped being a single-model autocomplete tool a long time ago. In chat and in agent modes, users can choose from several models, and that choice affects speed, answer quality and how fast usage limits are consumed. Being able to set a default model means you no longer have to pick one in every session. For a team, it is also a way to agree on a common standard: a fast model for routine work, a heavier one for hard tasks.

The second item concerns MCP, the Model Context Protocol. Agents use it to connect to external tools and data sources: an issue tracker, a knowledge base, an internal API, a database. The more such servers are connected, the more important it becomes to see which ones are active and what exactly the agent can call. The announcement promises "greater control" over the MCP server. Which specific switches and policies were added is not disclosed in the available text.

Diagnostics, chat and account

GitHub says diagnostics are now easier to "address." Judging by the wording, this refers to working with the warnings and errors the IDE highlights in code: you can pass them to Copilot faster and get fixes back. In JetBrains IDEs, code analysis is deeply built into the editor, and this is where an assistant can add value most easily without a long prompt.

Two more changes concern everyday ergonomics: chat navigation and account management. The first matters when a conversation grows to dozens of messages and you have to go back to earlier answers. The second is for people who work with several accounts, for example a personal one and a corporate one. GitHub does not specify what exactly changed in the visible part of the post.

What the announcement tells us

• Product: the GitHub Copilot plugin for JetBrains family IDEs
• Publication date: 10 October 2026
• Models: a setting for the default model
• Agents and tools: expanded control over the MCP server
• Code: easier work with diagnostics
• Interface: improved chat navigation, new account controls
• Not specified: version number, list of affected IDEs, pricing restrictions, and the items after the ellipsis

In scale, this is an incremental release. It matters not for new capabilities but because it shows that JetBrains remains a full platform for GitHub, not a second tier behind VS Code. There, Copilot competes with JetBrains' own AI assistant and with external agent tools, and parity on basic settings is a question of user retention for it.

What to check after updating

Update the plugin through the IDE's plugin manager and open the Copilot settings. Check which model is set as the default and, if needed, pin the one that fits your tasks and limits.

If you have connected MCP servers, go through the list: keep only the ones you need and make sure the access rights of each are clear. Organization administrators should check Copilot policies: corporate settings may restrict some of the new options.

Finally, try sending a real warning from your project to the chat and judge how accurate the fix is. Read the full list of changes in the original GitHub Changelog entry: it is cut off in the public announcement.