JetBrains MCP: How to Add a Server (2026)

To add an MCP server to a JetBrains IDE, open Settings → Tools → AI Assistant → Model Context Protocol, click Add, and paste a JSON config — a command for a local server, a url for a remote one. Because MCP lives in the shared AI Assistant plugin, the exact same steps work in IntelliJ IDEA, PyCharm, WebStorm, Rider, and the rest of the family. This guide walks through both server types, the DesignRevision MCP setup, and the one bit of "JetBrains MCP" that confuses everyone.

Last updated: July 2026. JetBrains' MCP support arrived recently (AI Assistant 2026.1); the flow below was verified against JetBrains' AI Assistant MCP docs in July 2026 — check there if a menu label has moved.

"JetBrains MCP" Means Two Things

Search "JetBrains MCP" and you'll hit two different features. Know which one you want:

1. JetBrains as an MCP client — adding MCP servers to AI Assistant so it can call external tools. That's this guide.
2. JetBrains as an MCP server — the bundled mcp-jetbrains plugin that exposes the IDE's own tools (file read/write, search, run configurations) to an outside client like Claude Desktop or Cursor. On by default since IDE version 2025.2; we cover it briefly at the end.

If you want your JetBrains IDE's AI to reach your repos, database, or component library, keep reading.

Prerequisites

• A JetBrains IDE with AI Assistant (2026.1 or later) — IntelliJ IDEA, PyCharm, WebStorm, Rider, GoLand, PhpStorm, RubyMine, CLion, or Android Studio. MCP is part of the AI Assistant plugin, so any of them works.
• The server details — a launch command for a local server, or an HTTPS URL for a remote one.

Step 1: Open the MCP Settings

Go to Settings → Tools → AI Assistant → Model Context Protocol (MCP). (You can also type / in the AI Assistant chat and choose the add-command option.) This page lists your configured servers and is where you add new ones.

Step 2: Add a Local (STDIO) Server

Click Add, choose STDIO, and provide a JSON config — the same mcpServers shape other clients use:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "~/code"]
    }
  }
}

Optionally set a working directory for relative paths, choose whether the server is global or project-only, then click OK and Apply. AI Assistant launches the server and its tools become available in chat.

Step 3: Add a Remote (HTTP/SSE) Server

For a hosted server, provide a url instead of a command:

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

AI Assistant supports streamable HTTP and SSE for remote servers. One caveat worth knowing up front: JetBrains' remote config doesn't document a place for custom auth headers, so a server that needs a bearer token won't authenticate through the plain url form — which is exactly the case handled next.

Add DesignRevision MCP to JetBrains

DesignRevision MCP — the server that installs real shadcn/ui components — is a remote server that authenticates with a bearer token. Because AI Assistant's remote url config has no header field, the reliable way to add it today is to bridge the remote endpoint into a local STDIO command with mcp-remote, which carries the token for you.

In Settings → Tools → AI Assistant → Model Context Protocol, add a STDIO server with this config:

{
  "mcpServers": {
    "design-revision": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://mcp.designrevision.com/mcp",
        "--header", "Authorization: Bearer YOUR_TOKEN"
      ]
    }
  }
}

Apply it, and AI Assistant can discover, inspect, and install shadcn/ui components on request. It's freemium, so you can wire it up and pull components on the free plan. This is the same bridge pattern used for Claude Desktop and Perplexity.

It Works the Same in Every JetBrains IDE

Because MCP is part of AI Assistant, not a per-IDE feature, the Settings path and JSON are identical whether you're in IntelliJ IDEA, PyCharm, WebStorm, Rider, GoLand, PhpStorm, RubyMine, CLion, or Android Studio. Configure a server once and the setup transfers; only the IDE around it changes. Servers set to global follow you across every JetBrains IDE on your machine.

Where JetBrains Keeps Your MCP Config

One difference from other clients is worth calling out. Cursor and Codex store MCP servers in a file you can commit — .cursor/mcp.json or ~/.codex/config.toml. JetBrains keeps yours in the IDE's settings, not a project file, and the global vs. project switch you pick when adding a server decides its reach: global servers follow you into every JetBrains IDE on the machine, while project servers stay with that one project. That's convenient for a personal setup but means there's no committable file to share a server across your team automatically — each teammate adds it through the same Settings panel. If you want a repo-level, shareable config, that's a point in favor of Cursor or Claude Code instead.

The Other Direction: JetBrains as an MCP Server

The flip side of JetBrains MCP is the mcp-jetbrains server: since IDE version 2025.2, JetBrains ships a bundled MCP server that lets an outside client — Claude Desktop, Cursor, VS Code — drive the IDE, running configurations and reading or editing files without leaving that client. If you searched "JetBrains MCP server" expecting to expose your IDE rather than extend it, that's the plugin you want. It's a different job from everything above, where AI Assistant is the client.

Troubleshooting

• Server won't start. For a STDIO server, run the exact command and args in a terminal first — if npx or the package name fails there, AI Assistant can't launch it either.
• Remote server added but returns 401. The plain url form can't send a token. Switch to the mcp-remote STDIO bridge above so the Authorization header goes through.
• Tools don't appear. Confirm the server is enabled and set to the right level (global vs project), then reopen the chat so the tool list refreshes.
• Wrong AI Assistant version. MCP needs AI Assistant 2026.1+. Update the plugin (and the IDE) if the Model Context Protocol page isn't there.

Which MCP Servers Should You Add?

The servers are the same across every agent — only the config changes. A focused set of GitHub, Context7, a database server, and DesignRevision MCP for UI covers most work. We rank the best picks in our roundups for Claude Code, Cursor, and Codex, and compare how each agent configures MCP in our MCP clients comparison.

Conclusion

JetBrains MCP is two features sharing a name: AI Assistant as a client (add servers in Settings → Tools → AI Assistant → Model Context Protocol) and the bundled IDE MCP server (expose the IDE to other clients). For the client side, local servers take a command, remote ones take a url, and a bearer-authed remote server like DesignRevision MCP goes through the mcp-remote bridge. Set it once and it works across IntelliJ IDEA, PyCharm, WebStorm, and every other JetBrains IDE.

Related Resources

• DesignRevision MCP — the shadcn/ui MCP server
• MCP Clients Compared: Every AI Coding Agent
• ChatGPT MCP: How to Add a Server (2026)
• Perplexity MCP: The Complete Guide (2026)
• How to Add an MCP Server to Claude Code (2026 Guide)
• Claude Desktop MCP Config: How to Add a Server (2026)
• Browse the shadcn/ui component registry