MCP Server Config Generator

MCP (Model Context Protocol) is an open standard developed by Anthropic that lets AI assistants like Claude connect to external tools, data sources, and services. Think of it as a universal plugin system for AI — instead of hardcoding integrations, MCP lets AI clients discover and call server-defined tools at runtime. This tool generates the JSON configuration that tells your AI client which MCP servers to load on startup.

On macOS, the file lives at ~/Library/Application Support/Claude/claude_desktop_config.json. On Windows, it's at %APPDATA%\Claude\claude_desktop_config.json (typically C:\Users\YourName\AppData\Roaming\Claude\). On Linux, check ~/.config/Claude/claude_desktop_config.json. You can also find it via Claude Desktop → Settings → Developer → Edit Config, which opens the file directly in your editor.

stdio (standard input/output) is the most common transport: the AI client launches the MCP server as a subprocess and communicates via stdin/stdout pipes. This works great for local servers installed via npm, Python, or other runtimes. SSE (Server-Sent Events) connects to a remote HTTP server instead — useful for shared servers, cloud-hosted MCP services, or servers that need to stay running independently. For local development, stdio is usually simpler; for production services, SSE is preferred.

In the stdio transport section, click '+ env var' to add key-value pairs. These become the env object in the JSON config, which the AI client passes to the server process as environment variables. This is the recommended way to provide API keys and secrets to MCP servers — never hardcode secrets in the args array. Example: set GITHUB_TOKEN as a key with your personal access token as the value for the GitHub MCP server.

The official Anthropic MCP repository includes servers for: filesystem (read/write local files), GitHub (repos, issues, PRs), PostgreSQL and SQLite (database queries), Brave Search and Fetch (web search), Memory (persistent key-value store), Puppeteer (browser automation), Slack (messaging), and Google Drive. The community has built hundreds more. Install most via npx -y @modelcontextprotocol/server-name. Browse the full list at github.com/modelcontextprotocol/servers.

Yes — switch the Target Client selector to generate config for Cursor (~/.cursor/mcp.json), Continue (~/.continue/config.json), or Windsurf (~/.codeium/windsurf/mcp_config.json). The mcpServers JSON structure is identical across all clients since they all follow the MCP specification. Only the file location differs. The 'File Path' tab shows where to save the config for each client and includes setup instructions.