MCP Config Generator
Generate mcpServers config for Claude Desktop, Cursor, and VS Code — visual editor, runs entirely in your browser
No env vars. Click "+ Add var" to add one.
Output format
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"unnamed": {
"command": "npx",
"args": []
}
}
}You’re setting up a new MCP server for Claude Desktop or Cursor and you’re tired of wrestling with JSON syntax errors and hunting down the right config file path. This tool lets you configure your MCP servers visually and instantly outputs valid JSON for Claude Desktop, VS Code, or Cursor — no manual editing required.
What Is the Model Context Protocol?
The Model Context Protocol (MCP) is an open standard introduced by Anthropic in late 2024 that lets AI assistants like Claude connect to external data sources, tools, and services. Think of it as a universal adapter that lets your AI client talk to databases, file systems, APIs, and custom servers through a standardized interface.
Before MCP, every AI integration required bespoke connectors — custom function calls, specialized plugins, or one-off API wrappers. MCP standardizes the communication layer so a single server implementation works with any MCP-compatible client: Claude Desktop, Cursor, VS Code Copilot, and an expanding ecosystem of tools.
How MCP Works
MCP follows a client-server architecture:
- MCP Host — the AI application (Claude Desktop, Cursor, etc.) that wants to use external capabilities
- MCP Client — lives inside the host and manages the connection to MCP servers
- MCP Server — a lightweight process that exposes resources (files, database records), tools (functions the AI can call), and prompts to the client
The host starts MCP servers as child processes using stdio transport (the most common setup) and communicates using a JSON-RPC 2.0 protocol. Your config file tells the host which servers to launch and how.
Common MCP Servers
| Server | Package | What It Provides |
|---|---|---|
| Filesystem | @modelcontextprotocol/server-filesystem | Read/write files within allowed directories |
| GitHub | @modelcontextprotocol/server-github | Search repos, read files, manage issues and PRs |
| PostgreSQL | @modelcontextprotocol/server-postgres | Query and inspect Postgres databases |
| Brave Search | @modelcontextprotocol/server-brave-search | Web and local search via Brave API |
| Memory | @modelcontextprotocol/server-memory | Persistent key-value memory across conversations |
| Puppeteer | @modelcontextprotocol/server-puppeteer | Browser automation and web scraping |
Setting Up MCP Step by Step
1. Install Node.js (version 18+). Most official MCP servers are distributed as npm packages and run via npx.
2. Configure your host. Use this generator to produce a valid config, then place it at the correct path:
- Claude Desktop (macOS):
~/Library/Application Support/Claude/claude_desktop_config.json - Claude Desktop (Windows):
%APPDATA%\Claude\claude_desktop_config.json - Cursor:
~/.cursor/mcp.jsonor through Cursor Settings → Features → MCP
3. Restart your host. Claude Desktop and Cursor read the config at startup. Changes require a restart to take effect.
4. Verify the server appears. In Claude Desktop, check the MCP icon in the bottom toolbar — active servers show there. In Cursor, open Settings → Features → MCP to see connected servers.
Understanding the Config Format
The Claude Desktop format wraps all servers under a mcpServers key:
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/projects"],
"env": {}
},
"github": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-github"],
"env": {
"GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_your_token_here"
}
}
}
}Key points:
command— the executable to run (usuallynpx,python, ornode)args— array of arguments passed to the command;-ytells npx to skip confirmation promptsenv— environment variables injected into the server process; keep secrets here instead of in args
Environment Variables and Security
Never put API keys or secrets directly in args. The env object is the right place — it keeps credentials out of process arguments, which can appear in system logs. For additional security, consider using a secrets manager and referencing environment variables that are already set in your shell profile.
The env object in MCP config merges with (not replaces) the existing environment, so system variables remain accessible.
Frequently Asked Questions
Why does my MCP server show as disconnected in Claude Desktop?
The most common causes are: (1) the command path is wrong or the package isn’t installed, (2) the allowed directory path in filesystem servers doesn’t exist, or (3) a required API key in env is missing or invalid. Check Claude Desktop logs at ~/Library/Logs/Claude/ (macOS) for detailed error messages.
Can I run multiple instances of the same MCP server?
Yes — give each instance a unique key in mcpServers. For example, you can have "projects-fs" and "downloads-fs" both pointing to the filesystem server but with different allowed directories.
What’s the difference between Claude Desktop and Cursor config formats?
They’re functionally identical — both use mcpServers as the top-level key. VS Code’s format wraps it under "mcp.servers" to follow VS Code’s namespaced settings convention. This generator handles the formatting difference automatically.
Do I need to install MCP servers globally?
No. Using npx -y <package> downloads and runs the server on-demand. If you prefer, you can install globally with npm install -g <package> and then use the bare package name as the command, but npx works fine and avoids polluting your global node_modules.