MCP_CONFIG.mdā¢1.74 kB
# MCP Server Configuration Examples
This document shows how to configure various AI assistants to use your MCP server.
## Claude Desktop Configuration
Add this to your Claude Desktop configuration file:
### Windows
Location: `%APPDATA%\Claude\claude_desktop_config.json`
### macOS
Location: `~/Library/Application Support/Claude/claude_desktop_config.json`
```json
{
"mcpServers": {
"sdk-ai-mcp": {
"command": "node",
"args": ["C:\\path\\to\\your\\project\\index.js"],
"env": {}
}
}
}
```
## Continue.dev Configuration
Add this to your Continue configuration:
```json
{
"mcpServers": [
{
"name": "sdk-ai-mcp",
"command": "node",
"args": ["./index.js"],
"cwd": "/path/to/your/project"
}
]
}
```
## Generic MCP Client Configuration
For any MCP-compatible client:
- **Command**: `node`
- **Arguments**: `["index.js"]`
- **Working Directory**: Path to this project
- **Communication**: stdio (standard input/output)
## Testing the Connection
1. Start your MCP client with the configuration above
2. The server should automatically connect via stdio
3. You can then use the available tools:
- `echo` - Echo back text
- `get_time` - Get current time
- `calculate` - Perform arithmetic
## Environment Variables
You can set these environment variables for additional configuration:
- `NODE_ENV` - Set to "development" for debug logging
- `MCP_SERVER_NAME` - Override the server name (default: "example-mcp-server")
## Troubleshooting
If the connection fails:
1. Check that Node.js is installed and accessible
2. Verify the path to `index.js` is correct
3. Ensure the MCP client supports stdio transport
4. Check the console output for error messages