Agent MCP

4/18/2025

Learn how to use popular MCP clients with different MCP servers to enhance your AI workflow.

How to Use MCP Clients with MCP Servers

Introduction

The Model Context Protocol (MCP) enables powerful integrations between AI assistants and external tools or data sources. This guide will walk you through how to use some of the most popular MCP-enabled clients with various MCP servers to enhance your workflow.

1. VS Code with MCP

Visual Studio Code offers MCP integration through extensions, allowing you to leverage AI assistance with access to your codebase and other tools.

Setting Up VS Code with MCP

  1. Install VS Code from code.visualstudio.com
  2. Install the GitHub Copilot extension which includes MCP support
  3. Configure MCP servers:
    • Open VS Code settings (Ctrl+,)
    • Search for "MCP"
    • Add your MCP server configurations

Installing Smithery.ai MCP Servers

Smithery.ai provides a variety of pre-built MCP servers. To set up in VS Code:

  1. Open the terminal (Ctrl+`) and run the following commands to install Smithery.ai servers:

    # Install Brave Search server
npm install -g @smithery-ai/brave-search-mcp

# Install GitHub server
npm install -g @smithery-ai/github-mcp

# Install Fetch server
npm install -g @smithery-ai/fetch-mcp

  2. Register the servers: After installation, you need to register the servers:

    # Using the MCP CLI tool to register
mcp register @smithery-ai/brave-search-mcp
mcp register @smithery-ai/github-mcp
mcp register @smithery-ai/fetch-mcp

  3. Configure authentication (if needed):

    • For GitHub server: 
      mcp config set @smithery-ai/github-mcp.accessToken YOUR_GITHUB_TOKEN
    • You can generate a personal access token from the GitHub settings page

Using MCP Servers in VS Code

Once configured, you can use MCP servers directly from the Copilot chat:

/mcp use @smithery-ai/github

This will activate the GitHub MCP server, allowing Copilot to interact with GitHub repositories, issues, and pull requests.

Common commands include:

  • Searching repositories: Search for repositories about MCP
  • Creating issues: Create an issue for this bug I found
  • Viewing pull requests: Show open PRs for this repository

Using Brave Search example:

/mcp use @smithery-ai/brave-search
Search for information about "Model Context Protocol"

Using Fetch example:

/mcp use @smithery-ai/fetch
Get user information from https://api.github.com/users/octocat

2. Cursor with MCP

Cursor is an AI-powered code editor built on VS Code that includes native MCP support for enhanced AI capabilities.

Setting Up Cursor with MCP

  1. Download and install Cursor from cursor.sh
  2. Enable MCP integration:
    • Go to Settings > AI Features
    • Enable "MCP Integration"
    • Add MCP servers in the provided configuration area

Installing Smithery.ai MCP Servers in Cursor

Cursor provides a simplified installation process for MCP servers:

  1. Open Cursor command palette (Ctrl+Shift+P), search for and select "MCP: Manage Servers"

  2. Add Smithery.ai servers:

    • Click "Add New Server"
    • In the URL field, enter Smithery.ai server addresses: 
      https://smithery.ai/server/@smithery-ai/brave-search
https://smithery.ai/server/@smithery-ai/github
https://smithery.ai/server/@microsoft/playwright

  3. Configure server parameters:

    • For servers requiring authentication, provide API keys or access tokens
    • For GitHub, you need to generate a personal access token
    • For Google Maps, you need to provide a valid API key

  4. Server priority settings:

    • Cursor allows you to set server priorities for which server to use in case of conflicts
    • Reorder servers using the drag-and-drop interface

  5. Test server connections:

    • Select a server and click the "Test Connection" button
    • Cursor will verify the server is working properly and display a list of capabilities

Using Cursor's MCP Control Panel

Cursor provides a dedicated MCP control panel that you can access by:

  1. Clicking the "MCP" icon in the sidebar
  2. Viewing the list of installed servers and their status
  3. Performing actions for each server:
    • Enable/disable
    • View documentation
    • Test connection
    • Edit configuration

Using MCP Servers in Cursor

Cursor's AI chat has built-in MCP support. You can:

  1. Access filesystem using the Filesystem MCP server:

    Could you analyze the files in my project directory?

  2. Run web searches with the Brave Search MCP server:

    Search for information about MCP protocol specifications

  3. Interact with GitHub using the GitHub MCP server:

    Clone repository modelcontextprotocol/python-sdk

  4. Use Google Maps:

    Calculate the driving distance from Beijing to Shanghai

  5. Automate browser tasks with Playwright:

    Visit wikipedia.org using Playwright and take a screenshot, then analyze the page structure

Cursor will automatically use the appropriate MCP server based on the task. You can also explicitly specify servers using the @ symbol in chat:

@github create a new issue report about MCP integration
@brave-search find the latest research on AI safety

3. Windsurf Editor

Windsurf is a dedicated AI programming assistant with comprehensive MCP integration.

Setting Up Windsurf with MCP

  1. Download and install Windsurf from windsurf.io
  2. Configure MCP servers:
    • Navigate to Preferences > Integrations > MCP
    • Click "Add Server" and enter the server details
    • Enable the servers you want to use

Installing Smithery.ai MCP Servers in Windsurf

Windsurf provides a visual interface for installing and managing MCP servers:

  1. Open the MCP Server Manager:

    • Click "Tools" in the top menu bar > "MCP Servers"
    • Or use the shortcut Alt+M

  2. Add Smithery.ai servers:

    • Click the "Add Server" button
    • Select Smithery.ai servers from the "Preset Servers" dropdown
    • Or manually enter server URLs: 
      https://smithery.ai/server/@smithery-ai/brave-search
https://smithery.ai/server/@smithery-ai/github
https://smithery.ai/server/@microsoft/playwright

  3. Configure server parameters:

    • For servers requiring authentication, provide API keys or access tokens
    • For GitHub, you need to generate a personal access token
    • For Google Maps, you need to provide a valid API key

  4. Server priority settings:

    • Windsurf allows you to set server priorities for which server to use in case of conflicts
    • Reorder servers using the drag-and-drop interface

  5. Test server connections:

    • Select a server and click the "Test Connection" button
    • Windsurf will verify the server is working properly and display a list of capabilities

Using MCP Servers in Windsurf

Windsurf's AI assistant panel provides a chat interface with MCP support:

  1. Access file system:

    Show me all JavaScript files in this project

  2. Run database queries using the PostgreSQL MCP server:

    Query user data from our database

  3. Automate browser tasks with the Playwright MCP server:

    Visit https://example.com and take a screenshot, then analyze the page structure

  4. Use Google Maps geographical services:

    Calculate the driving route from Guangzhou to Shenzhen, and estimate travel time

  5. Interact with GitHub:

    List the open issues in my GitHub repository and sort them by priority

Windsurf also provides a visual MCP server manager where you can see all connected servers and their status.

Windsurf's Unique MCP Features

Windsurf offers some unique MCP features not found in other clients:

  1. MCP Server Composition: Create custom server groups that function as a single server
  2. MCP Workflow Automation: Record and replay common MCP interaction sequences
  3. MCP Status Monitoring: Monitor MCP server performance and usage in real-time
  4. Offline MCP Caching: Cache MCP results for use when you don't have network connectivity

4. Claude Desktop

Claude Desktop is Anthropic's official desktop application for Claude, with native MCP support.

Setting Up Claude Desktop with MCP

  1. Download and install Claude Desktop from claude.ai/download

  2. Install MCP servers:

    • From a terminal, use the mcp install command: 
      mcp install <path-to-server-file>
    • Alternatively, use pre-built servers: 
      mcp install @smithery-ai/filesystem

  3. Restart Claude Desktop to apply changes

Using MCP Servers in Claude Desktop

Once installed, you can use MCP servers directly in conversation with Claude:

  1. Access files:

    Can you read the contents of my report.pdf file?

  2. Make API requests with the Fetch MCP server:

    Can you get the weather forecast for New York City?

  3. Use Google Maps:

    What's the distance between San Francisco and Los Angeles?

Claude will automatically leverage the appropriate MCP server based on your request.

5. Cline

Cline is a command-line interface for interacting with Claude and MCP servers.

Setting Up Cline with MCP

  1. Install Cline using npm:

    npm install -g cline-cli

  2. Configure your API key:

    cline config set apiKey YOUR_API_KEY

  3. Install MCP servers:

    cline mcp install @smithery-ai/github

Using MCP Servers in Cline

You can interact with MCP servers directly from the command line:

  1. Start a conversation:

    cline chat

  2. Use MCP capabilities:

    # In chat mode
> Can you clone the repository at https://github.com/example/repo?

Cline also supports scripting MCP interactions for automation.

6. Alibaba Cloud Bailian

Alibaba Cloud's Bailian platform includes MCP integration for enterprise AI applications.

Setting Up Bailian with MCP

  1. Create an Alibaba Cloud account and access the Bailian console
  2. Navigate to the MCP tab
  3. Configure MCP servers:
    • Click "Add Server"
    • Configure connection details for your MCP servers

Using MCP Servers in Bailian

Bailian provides a web interface for interacting with AI that leverages MCP:

  1. Access enterprise data sources:

    Analyze our sales data for Q1 2025

  2. Interact with cloud resources:

    Scale our Kubernetes cluster to handle increased traffic

  3. Connect to internal tools:

    Generate a report from our CRM system

Advanced MCP Usage Tips

Regardless of which client you use, these tips will help you get the most out of MCP:

1. Composing Multiple Servers

Most MCP clients allow you to use multiple servers together. For example:

Can you search for information about TensorFlow (using Brave Search), 
then create a Python file (using Filesystem) that implements a basic neural network?

2. Authentication

Some MCP servers require authentication. Make sure to:

  • Set up credentials in your client's MCP settings
  • Use token-based authentication where possible
  • Securely store your credentials

3. Error Handling

If an MCP server fails to respond:

  • Check if the server is running
  • Verify network connectivity
  • Look at server logs for error messages
  • Ensure you have the correct permissions

4. Performance Optimization

For better performance with MCP:

  • Install frequently used servers locally
  • Use proxy servers for remote MCP services
  • Cache results when appropriate

Conclusion

MCP is transforming how we interact with AI assistants by providing standardized access to external tools and data. By understanding how to configure and use MCP clients effectively, you can build powerful workflows that combine the intelligence of large language models with the capabilities of specialized tools.

Whether you're a developer, data scientist, or business user, MCP clients offer a way to enhance AI with real-world context and capabilities, making your interactions more productive and meaningful.

Back to articles