Article

Understanding default models and MCP startup controls in Copilot for JetBrains

Copilot for JetBrains adds controls for enterprise model defaults and automatic MCP startup. Understanding how these differ from access policies helps teams evaluate the update.

Share

Koharu's reading tip

Separate what starts by default from what is permitted. That makes it easier to decide which settings your team needs to align.

Koharu's reading tip

When a team uses GitHub Copilot, the model that starts a conversation and the moment external tools become active are useful settings to align. The October 10, 2026 update for JetBrains provides a reason to revisit that setup.

Choosing a default and granting access serve different purposes. Understanding their boundaries helps you decide where to look when a model or MCP tool is unavailable.

Model defaults set the starting point; access policies set availability

Enterprise administrators can now specify an available agent model for new conversations through managed settings, while explicit model-picker selections remain respected. Update announcement

As background, model availability depends on the subscription plan, client, and organization or enterprise restrictions. Enabling or disabling models belongs to model access configuration.

For a team recommendation, first establish whether the intended account can use that model in JetBrains, then align the initial selection. A default setting alone cannot explain why a model is absent from the picker.

During evaluation, test the initial selection in a new conversation separately from an explicit user selection. This gives you a concrete way to check the intended behavior.

MCP startup and access to connected services are separate controls

Once the model is selected, consider its external tools. MCP connects AI applications to external context and tools. Copilot for JetBrains supports both remote servers and local servers that run on your machine and can access local resources. Official MCP setup guide

The update adds an option to disable automatic MCP server startup for Copilot and Claude. Update announcement Evaluate this as a decision about when tools become active, separately from what they may access.

For example, the remote GitHub MCP server uses approved OAuth scopes or the scopes granted to a PAT, with organization restrictions also applying. These are authentication and authorization controls, separate from automatic startup. GitHub MCP server configuration

For users with organization-provided Copilot Business or Enterprise seats, the MCP servers in Copilot policy must also be enabled. If a tool is unavailable, investigate server startup, permission to use MCP, and access to the connected service separately.

Test conversation and tool startup on JetBrains IDE 2025.2 or later

The Copilot plugin has ended support for JetBrains IDE 2025.1 and requires 2025.2 or later. Update announcement Environments pinned to 2025.1 need to assess the IDE upgrade first.

A useful evaluation sequence is to choose the account, test model selection in a new conversation, and then check the required MCP tools. Changing one part at a time makes the intended controls easier to assess.

Team procedures should distinguish conversation and tool startup conditions from permission to use them. Keeping those responsibilities explicit lets you adopt the new controls without mistaking defaults or startup settings for complete access management.

Source

Share

Related Articles

These articles share nearby categories or tags, so you can keep reading along the same thread.