Claude Desktop
Anthropic's Claude with Model Context Protocol support. Excellent for conversational workflows, analysis, and exploration.
Welcome to the guide for using Claude Desktop with MCP servers! This document shows you how to connect Claude Desktop to Model Context Protocol servers, enabling powerful AI-assisted workflows.
Info Info
Experimental: Claude Desktop MCP integration is evolving. Cursor IDE is currently more mature for development workflows, but Claude Desktop excels at conversational exploration and analysis.
Overview
Claude Desktop is Anthropic’s native desktop application that supports the Model Context Protocol (MCP), allowing Claude to interact with external systems. This integration provides:
- MCP Server Access: Connect to custom MCP tools and resources
- Conversational Workflows: Natural back-and-forth for complex tasks
- Tool Integration: Use specialized tools through MCP protocol
- Data Exploration: Query and analyze data using natural language
Part 1: Installation
Step 1: Get Claude Desktop
Download Claude Desktop
- Visit claude.ai/download
- Choose your operating system (macOS or Windows)
- Download and install the application
Sign In
- Launch Claude Desktop
- Sign in with your Anthropic account
- Claude Pro subscription recommended for heavy usage
Verify Installation
- Open Claude Desktop
- Test with a simple query: “Hello, what can you help me with?”
Part 2: Configuring MCP Servers
Locate Configuration File
macOS:
~/Library/Application Support/Claude/claude_desktop_config.json
Windows:
%APPDATA%\Claude\claude_desktop_config.json
Configure Your MCP Server
Edit the configuration file:
{
"mcpServers": {
"my-mcp-server": {
"command": "node",
"args": ["/path/to/your/mcp-server/build/index.js"],
"env": {
"API_KEY": "your-api-key-here"
}
}
}
}
For HTTP-based MCP servers:
{
"mcpServers": {
"my-http-server": {
"url": "http://localhost:3000/mcp",
"transport": "http"
}
}
}
Tip Tip
Multiple Servers: You can connect to multiple MCP servers by adding additional entries in the mcpServers object. Claude will have access to tools from all connected servers.
Restart Claude Desktop
After saving the configuration:
- Completely quit Claude Desktop (Cmd+Q on Mac, Alt+F4 on Windows)
- Restart the application
- Claude will connect to your configured MCP servers on startup
Part 3: Using Claude with MCP Servers
Verify Connection
Ask Claude about available tools:
“What tools do you have access to?”
Claude should list the tools from your connected MCP servers.
Example Workflows
Data Exploration:
“Can you query the database and show me the top 10 customers by revenue?”
Code Generation:
“Help me write a Python function that processes this data using the available tools”
Documentation:
“Explain how the authentication system works in this codebase”
Best Practices
- Be Specific: Clearly describe what you want Claude to do
- Reference Tools: Mention specific tool names when you know them
- Iterate: Start with simple requests, then refine
- Review Output: Always verify Claude’s responses and generated code
Part 4: Troubleshooting
MCP Server Not Connecting
Symptoms: No tools available, connection errors
Solutions:
- Verify MCP server is running (
curl http://localhost:3000/health) - Check configuration file path is correct
- Review server logs for connection attempts
- Restart Claude Desktop completely
Tools Not Working
Symptoms: Claude can see tools but can’t execute them
Solutions:
- Check MCP server logs for errors
- Verify authentication/API keys are correct
- Ensure server has proper permissions
- Test tools directly with MCP client
Configuration Not Loading
Symptoms: Changes to config file don’t take effect
Solutions:
- Verify JSON syntax is valid (use a JSON validator)
- Check file path is exactly correct for your OS
- Completely quit and restart Claude Desktop (not just close window)
- Check Claude Desktop logs for errors
Part 5: Advanced Configuration
Environment Variables
Pass environment variables to your MCP server:
{
"mcpServers": {
"production-server": {
"command": "node",
"args": ["server.js"],
"env": {
"NODE_ENV": "production",
"API_KEY": "${API_KEY}",
"DATABASE_URL": "${DATABASE_URL}"
}
}
}
}
Multiple Servers
Connect to multiple MCP servers simultaneously:
{
"mcpServers": {
"database-server": {
"url": "http://localhost:3000/mcp"
},
"file-server": {
"url": "http://localhost:3001/mcp"
},
"api-server": {
"url": "http://localhost:3002/mcp"
}
}
}
Claude will have access to tools from all servers!
Comparison: Claude Desktop vs Cursor
| Feature | Claude Desktop | Cursor IDE |
|---|---|---|
| Best For | Conversational exploration | Active development |
| Interface | Chat-focused | Code editor |
| MCP Support | Native | Native |
| Code Editing | Copy/paste | Inline editing |
| Codebase Context | Limited | Full workspace |
| Use Case | Analysis, planning | Writing code |
Tip Tip
Use Both: Many developers use Cursor for coding and Claude Desktop for planning, analysis, and exploration. They complement each other well!
Next Steps
- Build an MCP server: MCP Server Template
- Try Cursor too: Cursor Setup Guide
- Join the community: GitHub Discussions
For questions or issues, visit GitHub Issues.