JetBrains MCP Setup: Add Servers to AI Assistant and Use the IDE's Built-In MCP Server

JetBrains IDEs such as IntelliJ IDEA and PyCharm work with MCP in two directions. As a client, AI Assistant can use outside MCP servers: go to Settings | Tools | AI Assistant | Model Context Protocol (MCP), click Add and paste a JSON config. As a server, the IDE itself can expose its tools to agents like Claude Code or Codex: go to Settings | Tools | MCP Server and tick Enable MCP Server.

Most people need the first. Developers who run a terminal agent next to their IDE often want the second. This guide covers both, with the exact JSON format AI Assistant accepts.

Checked against: JetBrains' AI Assistant MCP documentation on 2 November 2026, plus GitHub's Copilot MCP docs for the Copilot plugin. eSEOspace did not test these products.

Part 1: Add an MCP server to AI Assistant

1. Open Settings | Tools | AI Assistant | Model Context Protocol (MCP). You can also type / in the AI chat and choose the Add Command option.
2. Click Add.
3. Paste a JSON configuration (formats below).
4. Choose whether the server is available globally or only in the current project. You can change this later in the Level column.
5. Set the working directory if the server uses relative paths. JetBrains says this is "the path to the folder from which the server is launched."
6. Click OK and Apply.

JSON for a local (stdio) server

JetBrains' docs give this template:

{
  "mcpServers": {
    "yourServerName": {
      "command": "path-or-command-to-start-server",
      "args": ["optional-arguments-passed-to-server"]
    }
  }
}

JSON for a remote server

{
  "mcpServers": {
    "yourServerName": {
      "url": "https://example.com/mcp"
    }
  }
}

The docs say "the url parameter represents the HTTP endpoint of the MCP server." Both templates were checked against JetBrains' docs on 2 November 2026. The free MCP config generator can fill these in for a specific server.

Import from Claude Desktop

If you already set servers up in Claude Desktop, use Import from Claude on the same settings page. JetBrains says it adds all of that app's servers at once. The MCP project's local server guide lists where that file lives on each OS, and our Claude Desktop config file guide covers it in detail.

Check the connection

JetBrains says "you can monitor the status of your connection in the Status column." Click the icon there to see the server's tools. To have new or edited servers start on their own, enable Automatically enable new and changed MCP servers.

Part 2: Turn on the IDE's built-in MCP server

JetBrains' docs say "starting with version 2025.2, JetBrains IDEs come with an integrated MCP Server." It lets external agents use IDE features. Steps:

1. Open Settings | Tools | MCP Server.
2. Select Enable MCP Server.
3. Review the access in the "Enable MCP Server?" dialog and click Enable.
4. Click Apply.

To connect a client, click Auto-Configure next to it. JetBrains lists detected clients "like Junie, VS Code, Claude Code, Codex, Air, and GitHub Copilot CLI." Restart the client afterwards. For other clients, Copy Config gives you a block to paste by hand.

Our JetBrains MCP server profile covers what the server exposes and its security trade-offs.

Limit what agents can do

Two settings matter here:

• Exposed Tools: clear the Enabled checkbox for tools you do not want clients to use.
• Brave mode: the option "Run shell commands or run configurations without confirmation (brave mode)" lets clients run terminal commands without asking. Leave it off unless you fully trust every connected agent.

Part 3: GitHub Copilot in JetBrains

The GitHub Copilot plugin for JetBrains has its own MCP setup, separate from AI Assistant. GitHub's docs say to open Copilot Chat, switch to Agent mode and choose Add MCP Tools, which opens an mcp.json file. GitHub's examples use a top-level servers key, for example:

{
  "servers": {
    "memory": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-memory"]
    }
  }
}

For the remote GitHub MCP server in JetBrains, GitHub's docs use a personal access token in a requestInit header block. On Copilot Business or Enterprise, the "MCP servers in Copilot" policy is disabled by default. Our GitHub Copilot MCP setup guide covers that policy.

Kiro inside JetBrains

If your team moved from Amazon Q Developer, Kiro's migration guide says Kiro works inside JetBrains IDEs through AI Assistant via the Agent Client Protocol (ACP), after you install and sign in to the Kiro CLI. Our Kiro and Amazon Q MCP setup guide covers its MCP files.

Common mistakes

• Mixing up the two settings pages. AI Assistant's MCP page adds servers to the IDE. The Tools | MCP Server page exposes the IDE to other agents.
• Using servers in AI Assistant. AI Assistant's template uses mcpServers; Copilot's uses servers.
• Relative paths without a working directory. Set the working directory so arguments resolve.
• Turning on brave mode for convenience. It removes confirmation for shell commands.

Next step

Choose your first servers with our best MCP servers for developers list. If you run Claude Code next to IntelliJ, pair this with our claude mcp add guide. For help planning tool access across a team, see MCP development.

Frequently asked questions