Copilot for JetBrains requires 2025.2+: choosing default models and MCP startup

Check your IDE first: support for 2025.1 has ended

GitHub states in its October 10 announcement that Copilot support for JetBrains IDE 2025.1 has ended. The plugin requires 2025.2 or later. This is an announced version requirement; the update also introduces controls for default models and MCP startup.

First check the version of the IDE you are using. If it is still 2025.1, plan an upgrade to a supported version, then check for a Copilot plugin update instead of repeatedly reinstalling the plugin. The announcement does not list an exact minimum plugin version for these features.

GitHub links to the plugin versions page in the announcement; it is included in the sources below. Check the IDE requirement first, then the plugin version you intend to install. This separates an unsupported older environment from a new setting that is not yet visible.

Will an enterprise default replace the model I chose?

Enterprise administrators can now use managed settings to specify any available Copilot agent model as the default for new conversations. The key point is the starting choice for a new conversation. The announcement does not say that every conversation already in progress will switch to the same model.

An explicit user choice in the model picker is preserved. An administrator setting a default and a user actively choosing a model are therefore separate decisions. This announcement does not expand the set of available models, and it does not list additional fees, regions, or tenant rollout times.

In a hypothetical example, a team wants new conversations to start with one model, while a developer has explicitly selected another. The administrator default should not be interpreted as erasing that choice. This illustrates the rule; we have not carried it out in a real team. Check the model shown in the actual conversation.

Does disabling MCP startup disable the tools?

MCP lets AI clients connect to configured tool servers. The new setting can disable automatic server startup for Copilot and Claude, giving users more control over when those tools become active.

Disabling automatic startup changes startup behavior. It does not establish that a server has been uninstalled, that tools are permanently disabled, or that existing permissions have been revoked. The announcement does not provide an exact setting key or click path, so avoid changing configuration based on a guessed parameter.

If you need this control, inspect the MCP settings available in your current plugin and check the automatic-startup choice and the status of configured tools. Decide whether to enable tools for your work; do not grant unnecessary permissions simply to check the change.

What does Fix do next to a diagnostic?

Diagnostic intention menus now include Fix. Selecting it opens inline chat and asks Copilot to repair the issue. It uses agent mode when that mode is available, and ask mode otherwise. This does not promise agent capabilities for every account.

The choice between the two modes depends on current availability. Entering ask mode alone does not establish that the update failed. Interpret the diagnostic in your project context, and review any changes proposed by the AI.

For a diagnostic you need to address, first read the error and identify the files involved. Then use Fix to inspect the suggestion. Check the scope and reasoning before deciding to apply it. The announcement describes a shorter route to a proposed fix, not a guarantee of an automatically correct repair.

What else changes in accounts and chat navigation?

Signed-in Copilot accounts now have clearer labels and status displays, improved navigation for long lists, and a Switch Account action. Before switching, confirm which account you are using. A clearer interface does not itself establish a change to permissions or billing rules.

Chat messages are grouped by turn, with controls for moving between turns. You can also copy or edit your own messages, making it easier to revisit the context you supplied and refine an unclear question.

New users get a one-time reminder of the Chat entry point. The sign-in page has improved layout, provider choices, and cancellable progress during authorization. The announcement also describes reliability improvements for model switching, MCP configuration, and worktrees, without guaranteeing the same outcome in every project.

What should you check after upgrading, and how can you report a problem?

First, confirm that your IDE is 2025.2 or later, and check the updates offered on the plugin versions page. Record the IDE and plugin versions separately so a later report can identify which layer is involved.

Second, inspect the current model in a new conversation. If your enterprise sets a starting default, distinguish that from an explicit model choice you made. When controlling tool startup, check the MCP setting and tool status separately; one switch does not explain every behavior.

Third, if a problem remains, record the specific symptoms, where they occur, and the IDE and plugin versions. The announcement points to in-IDE feedback and the GitHub repository for JetBrains plugin issues. The original links are retained below.

Include only the information needed for troubleshooting, keeping tokens, private code, and sensitive conversations out of feedback. Describe Fix using ask mode, a model missing from the available choices, and MCP not starting automatically as separate symptoms instead of calling them all a broken plugin.