MCP Client Setup¶

This page shows how to connect idfkit-mcp to every major MCP client. All clients run the same underlying command — only the config file location differs.

Common Command¶

All clients ultimately need to run:

idfkit-mcp

or

uvx --from idfkit-mcp idfkit-mcp

Use uvx when you want a managed, versioned runtime without installing into a project environment.

Transport Options¶

• Local desktop clients (Claude Desktop, Cursor, VS Code, etc.): use stdio (default).
• Hosted deployments: use streamable-http and expose a reachable host/port.

Claude Desktop¶

Claude Desktop was the first MCP client and remains one of the most popular.

Config file:

~/Library/Application Support/Claude/claude_desktop_config.json

%APPDATA%\Claude\claude_desktop_config.json

~/.config/Claude/claude_desktop_config.json

Example:

{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Restart Claude Desktop after saving.

Cursor¶

Cursor supports MCP servers via a global config file.

Config file:

~/.cursor/mcp.json

Example:

{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

VS Code¶

VS Code supports MCP servers through workspace settings or a project-level config file.

Workspace settings (settings.json):

{
  "mcp.servers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Or create a .vscode/mcp.json in your project root:

{
  "servers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Claude Code (CLI)¶

Claude Code is Anthropic's CLI tool. MCP servers can be configured globally or per-project.

Add via CLI:

claude mcp add idfkit -- uvx --from idfkit-mcp idfkit-mcp

Or edit the config files directly.

Project-level config (.mcp.json in project root — version-controllable):

{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Global config (~/.claude.json):

{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Verify with claude mcp list or type /mcp inside a session.

Windsurf¶

Windsurf (by Codeium) uses the same JSON format as Cursor.

Config file:

~/.codeium/windsurf/mcp_config.json

Example:

{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Restart Windsurf after saving.

ChatGPT Desktop¶

ChatGPT supports MCP servers via a local config file.

Config file:

~/Library/Application Support/com.openai.chat/mcp.json

%APPDATA%\com.openai.chat\mcp.json

Example:

{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Restart ChatGPT after saving.

Codex¶

In Codex, add a new MCP server pointing to idfkit-mcp (or the uvx wrapper command).

Recommended command:

uvx --from idfkit-mcp idfkit-mcp

Recommended working directory:

• The repository where IDF/epJSON files and simulation outputs should live.

Recommended prompt behavior:

• Start sessions by reading the idfkit://model/summary resource.
• Call describe_object_type before object creation or updates.
• Validate after every edit batch.

JetBrains IDEs¶

IntelliJ IDEA, PyCharm, WebStorm, and other JetBrains IDEs (2025.1+) support MCP servers natively.

1. Open Settings > Tools > AI Assistant > MCP Servers.
2. Click + to add a new server.
3. Set the command to uvx and arguments to --from idfkit-mcp idfkit-mcp.

Alternatively, add to the project-level .idea/mcpServers.json:

{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Cline¶

Cline is a VS Code extension with its own MCP settings panel.

1. Open the Cline sidebar in VS Code.
2. Click the MCP Servers icon (plug icon).
3. Click Configure MCP Servers to open cline_mcp_settings.json.

Example:

{
  "mcpServers": {
    "idfkit": {
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  }
}

Save the file and Cline will detect the new server automatically.

Continue¶

Continue is a VS Code / JetBrains extension for AI-assisted coding.

Config file:

~/.continue/config.yaml

Example (YAML):

mcpServers:
  - name: idfkit
    command: uvx
    args:
      - "--from"
      - idfkit-mcp
      - idfkit-mcp

Alternatively, in ~/.continue/config.json:

{
  "mcpServers": [
    {
      "name": "idfkit",
      "command": "uvx",
      "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
    }
  ]
}

Zed¶

Zed supports MCP servers via its settings file.

Open Zed > Settings (or ~/.config/zed/settings.json) and add:

{
  "context_servers": {
    "idfkit": {
      "command": {
        "path": "uvx",
        "args": ["--from", "idfkit-mcp", "idfkit-mcp"]
      }
    }
  }
}

Operational Tips¶

• Prefer absolute paths when loading or saving models.
• Keep one modeling task per server session to avoid state confusion.
• Use batch_add_objects when agents need to create many objects.
• Capture outputs in files if your client truncates long tool responses.