Copilot for JetBrains Adds Model and MCP Controls

GitHub has announced a group of controls and usability improvements for Copilot in JetBrains IDEs. The October 10, 2026 update covers model selection, diagnostic repair actions, Model Context Protocol (MCP) server startup, account management, and navigation within long Chat sessions.

The changes are relevant to both individual developers and organizations managing Copilot at scale. Administrators can define a preferred model for new conversations, while users retain a per-conversation model choice. Developers also get a faster path from an IDE diagnostic to a Copilot repair request. However, availability still depends on the Copilot plan, client support, and organization or enterprise policies, and GitHub’s supplied announcement does not establish one stable plugin version containing the entire feature set.

What changed in Copilot for JetBrains

Model defaults for managed environments

Enterprise administrators can set a preferred Copilot model for new conversations using the model key in managed-settings.json. The managed-settings documentation identifies that file as the usual location for managed configuration.

This setting establishes a default rather than permanently locking every conversation to one model. Users can still select another available model for an individual conversation. Which models appear in JetBrains depends on the user’s Copilot plan, the client, and any organization or enterprise restrictions. GitHub’s JetBrains model-selection documentation and supported-model guidance should therefore be treated as the source of truth for a particular environment.

The supplied documentation does not fully describe how this setting takes precedence across every management layer in JetBrains. Administrators should avoid assuming that a default model setting overrides all other policies or client-level behavior.

Diagnostic repair through inline Chat

The update adds a Fix action for diagnostics. When an IDE reports an issue, the action opens inline Chat and asks Copilot to repair it. GitHub says the action uses agent mode when that mode is available and falls back to ask mode otherwise.

That distinction matters operationally. The action is a shorter route to a repair request, but its behavior can vary with the modes enabled for the account and client. The feature should be considered an assistance workflow, not a substitute for reviewing the proposed change and validating the resulting code.

More control over MCP server startup

GitHub also announced a control for disabling automatic startup of configured MCP servers used by Copilot and Claude. MCP servers can expose external tools and data to an AI client, so controlling when they start can be useful for developers who do not want every configured integration to activate immediately.

JetBrains users configure MCP through the GitHub Copilot settings: open the Copilot menu, choose Edit Settings, go to Model Context Protocol, and select Configure. The configuration is stored in mcp.json. GitHub’s JetBrains MCP installation guide documents a remote-server setup using https://api.githubcopilot.com/mcp/ and a Bearer personal access token. It also notes that changing the configuration can restart defined MCP servers.

The exact user-facing name, syntax, scope, and precedence of the new automatic-startup setting are not documented in the supplied sources. Teams should use the control exposed by their installed client rather than adding an undocumented key to mcp.json.

MCP access is also subject to policy. For organization and enterprise users on Copilot Business or Copilot Enterprise, administrators can govern which MCP servers are permitted. The MCP management documentation describes allowlist and denylist controls. Those policy settings are separate from the newly announced automatic-startup control. The supplied documentation states that this organization and enterprise MCP policy does not govern individual Free, Pro, Pro+, and Max plans.

Chat, account, and sign-in improvements

The update includes several smaller changes aimed at making Copilot easier to operate in a busy JetBrains workspace:

• Signed-in account labels and status information are clearer.
• Long account lists have improved navigation, and a Switch Account action is available.
• Chat displays a one-time reminder, with revised sign-in headings and layout.
• Provider choices are clearer, and a direct link to Copilot plans is available.
• Browser-authorization progress can be canceled.
• Chat messages are grouped into turns with navigation controls.
• Users can copy and edit their own messages.
• The agent placeholder has been updated.
• Migration-session behavior now includes a Go Back action.

GitHub also lists reliability and display fixes across several workflows. These changes are less about adding a new programming capability than reducing friction when developers move between accounts, review a long conversation, or recover from an interrupted sign-in or migration flow.

Who is affected

The announcement requires JetBrains IDE 2025.2 or later; JetBrains IDE 2025.1 is no longer supported. Developers on an older IDE should upgrade before troubleshooting missing controls or Chat behavior.

Users also need access to GitHub Copilot to use Copilot Chat and related features. Model availability is not universal: it varies by plan, client, and organization or enterprise restrictions. GitHub’s supported-models documentation and feature matrix show that JetBrains support differs by extension version and that some capabilities are supported while others remain in preview or are unsupported.

The complete feature set’s stable Marketplace release has not been confirmed by the supplied evidence. The JetBrains Marketplace version history and feature-matrix information do not provide a single, definitive stable-version record for every item in the October 10 announcement. Availability may also differ by account or organization, so the changelog should not be read as a guarantee of simultaneous enablement for every Copilot user.

What developers and administrators should do

1. Check the IDE version. Confirm that affected installations use JetBrains IDE 2025.2 or later.
2. Verify the Copilot plugin. Update or inspect the installed plugin, but do not assume that one stable release contains every announced change.
3. Review model policy. Enterprise administrators should examine the model managed setting, available models, and applicable organization policies before presenting a model as the default.
4. Test the per-conversation picker. A managed default should establish the starting model for new conversations, while users may still need to choose another model when it is available and permitted.
5. Review MCP governance. Teams using MCP should inspect their mcp.json configuration and confirm that the required servers are allowed by organization or enterprise policy.
6. Handle tokens carefully. The documented remote GitHub MCP setup uses a personal access token because OAuth is stated as unsupported in that guide. Token storage, access, and scope should be treated as operational security concerns.
7. Use the diagnostic Fix action selectively. Check whether agent mode is available, review Copilot’s proposed changes, and validate the repair in the normal development workflow.

Availability caveats

GitHub’s update describes the new controls and improvements, but the supplied evidence does not establish a feature-by-feature rollout schedule or confirm that every account, plan, and organization received the changes simultaneously. The feature matrix also distinguishes supported, preview, and unsupported capabilities.

For MCP in particular, the startup control is announced, but its detailed configuration remains undocumented in the available material. Administrators should rely on the settings exposed by their current JetBrains installation and avoid publishing or enforcing an assumed configuration key.

Sources

• GitHub Changelog: New controls and chat improvements in Copilot for JetBrains
• GitHub documentation: Enterprise-managed settings
• GitHub documentation: Change the Chat model in JetBrains
• GitHub documentation: Supported models
• GitHub documentation: Copilot feature matrix
• GitHub documentation: MCP management
• GitHub MCP Server: Installation in other Copilot IDEs
• JetBrains Marketplace: GitHub Copilot plugin versions