Skip to main content

​
Installing mcp-use

Prerequisites:
  • Install Node.js (version 18 or higher)
  • npm, pnpm, or yarn package manager

​
Option 1: Create an MCP Server (for Building Servers)

The fastest way to scaffold a new MCP server project is to use create-mcp-use-app: 
npx create-mcp-use-app my-mcp-server
cd my-mcp-server
npm run dev
This command will:
  • Create a new directory with your project name
  • Initialize a TypeScript project with mcp-use configured
  • Set up a basic MCP server template with example tools
  • Install all dependencies automatically
Check out UI Widgets with support to MCP-UI and OpenAI Apps SDK for ChatGPT.

​
Option 2: Install as a Library (for MCP Agents)

Install mcp_use using your preferred package manager: 
npm install mcp-use
Then install your preferred LangChain provider: 
npm install @langchain/openai dotenv

​
Setting Up Your First Project

1

Create your MCP server or install the library

Use create-mcp-use-app for MCP servers or npm install for agents (see above).
2

Configure your LLM provider

Choose and install your preferred LangChain provider.
3

Set up environment variables

Create a .env file for your API keys:
.env
OPENAI_API_KEY=your_openai_key_here
ANTHROPIC_API_KEY=your_anthropic_key_here
GROQ_API_KEY=your_groq_key_here
GOOGLE_API_KEY=your_google_key_here
4

Verify installation

Test your installation with a simple script:
import { MCPAgent, MCPClient } from 'mcp-use'
console.log('mcp-use installed successfully!')

​
Development Installation

If you want to contribute or use the latest features, install from source: 
git clone https://github.com/mcp-use/mcp-use.git
cd mcp-use/libraries/typescript
pnpm install
pnpm run build

​
Installing MCP Servers

mcp-use connects to MCP servers that provide the actual tools.
Check out the Awesome MCP Servers repository for a comprehensive list of available servers.

​
Environment Setup

​
Environment Variables

Create a .env file in your project root:
.env
# LLM Provider Keys
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GROQ_API_KEY=gsk_...
GOOGLE_API_KEY=AI...

# Optional: Logging level
LOG_LEVEL=INFO

# Optional: MCP server paths
MCP_SERVER_PATH=/path/to/mcp/servers
Load environment variables in your scripts: 
import { config } from 'dotenv'

config()

// Your API keys are now available as environment variables
const openaiKey = process.env.OPENAI_API_KEY

​
Verification

Verify your installation works correctly: 
import { config } from 'dotenv'
import { ChatOpenAI } from '@langchain/openai'
import { MCPAgent, MCPClient } from 'mcp-use'

async function verifyInstallation() {
    config()

    // Simple configuration for testing
    const configuration = {
        mcpServers: {
            test: {
                command: 'echo',
                args: ['Hello from MCP!']
            }
        }
    }

    try {
        const client = new MCPClient(configuration)
        console.log('✅ MCPClient created successfully')

        const llm = new ChatOpenAI({ model: 'gpt-3.5-turbo' })
        console.log('✅ LLM initialized successfully')

        const agent = new MCPAgent({ llm, client })
        console.log('✅ MCPAgent created successfully')

        console.log('\n🎉 Installation verified! You\'re ready to use mcp_use.')

        // Clean up
        await client.closeAllSessions()

    } catch (error) {
        console.error('❌ Verification failed:', error)
        console.error('Please check your installation and API keys.')
    }
}

verifyInstallation().catch(console.error)

​
Next Steps

Quick Start

Follow our quickstart guide to build your first agent

Configuration

Learn how to configure MCP servers

LLM Integration

Explore different LLM providers and their setup

Examples

Browse real-world examples and use cases

​
Troubleshooting

Make sure mcp_use is installed and accessible:
npm list mcp-use
Verify your .env file is in the correct location and your API keys are valid:
cat .env  # Check file contents
Ensure your MCP servers are properly installed and accessible:
which npx  # For Node.js-based servers
Tool Calling Required: Only models with tool calling capabilities can be used with mcp_use. Make sure your chosen model supports function calling or tool use.