Skip to main content

​
Overview

MCP-use provides multiple deployment options to host your MCP servers in production. Whether you need a fully managed cloud solution or prefer self-hosting, mcp-use supports various deployment strategies to fit your needs.

​
Deployment Options

Manufact Cloud

Deploy to managed infrastructure with automatic scaling, observability, and GitHub integration

Self-Hosted

Deploy your MCP servers on your own infrastructure with full control

Docker

Containerize your MCP servers for consistent deployments across environments

Custom Clients

Integrate MCP servers with custom client applications

​
AI Assistant Integration

ChatGPT

Connect your MCP servers to ChatGPT for enhanced AI capabilities

Claude

Integrate with Claude Desktop for seamless AI interactions

​
Choosing a Deployment Strategy

Best for:
  • Quick production deployments
  • Teams wanting managed infrastructure
  • Projects requiring observability and metrics
  • GitHub-based workflows
Key Features:
  • Automatic deployments from GitHub
  • Branch-based deployments
  • Built-in observability and logging
  • Custom domains
  • Zero infrastructure management

​
Self-Hosted

Best for:
  • Organizations with strict data requirements
  • Custom infrastructure needs
  • Air-gapped environments
  • Complete control over resources
Key Features:
  • Full control over infrastructure
  • Custom networking configurations
  • On-premises deployment options
  • Integration with existing DevOps pipelines

​
Docker

Best for:
  • Consistent environments across dev/staging/prod
  • Kubernetes deployments
  • Multi-cloud strategies
  • CI/CD pipelines
Key Features:
  • Containerized deployments
  • Easy scaling and orchestration
  • Environment isolation
  • Portable across cloud providers

​
Architecture Considerations

​
Network Requirements

MCP servers communicate over HTTP/HTTPS using the Model Context Protocol. Ensure your deployment environment allows:
  • Inbound HTTP/HTTPS traffic on your configured port (default: 3000)
  • Outbound HTTPS for external API calls (if your tools require them)
  • WebSocket support for SSE (Server-Sent Events) streaming

​
Resource Requirements

Minimum requirements vary based on your server’s complexity and expected load.
Typical MCP Server:
  • CPU: 0.5-1 vCPU
  • Memory: 512MB-1GB RAM
  • Storage: 1-5GB (depends on widgets and assets)
High-Traffic MCP Server:
  • CPU: 2+ vCPUs
  • Memory: 2GB+ RAM
  • Storage: 5-10GB

​
Environment Variables

Most MCP deployments require configuration via environment variables: 
# Common environment variables
NODE_ENV=production
PORT=3000
MCP_SERVER_URL=https://your-server.run.mcp-use.com/mcp

# API keys and secrets (tool-specific)
OPENAI_API_KEY=sk-...
WORKOS_CLIENT_ID=client_...
WORKOS_API_KEY=sk-...
Never commit API keys or secrets to version control. Use environment variables or secret management systems.

​
Security Best Practices

​
Authentication

  1. OAuth Integration: Use OAuth providers (WorkOS, Auth0, Supabase) for user authentication
  2. API Keys: Secure your MCP endpoints with API key authentication
  3. HTTPS Only: Always use HTTPS in production

​
Secret Management

# Set environment variables during deployment
mcp-use deploy --env OPENAI_API_KEY=sk-...

# Or use .env file
mcp-use deploy --env-file .env.production

​
Network Security

  • Firewall Rules: Restrict access to trusted IP ranges
  • Rate Limiting: Implement rate limiting to prevent abuse
  • CORS Configuration: Configure CORS headers appropriately

​
Monitoring and Observability

​
Built-in Tools

MCP-use includes built-in observability features: 
import { MCPServer } from "mcp-use/server";
import { Langfuse } from "langfuse";

const server = new MCPServer({
  name: "my-server",
  version: "1.0.0",
  observability: {
    langfuse: new Langfuse({
      secretKey: process.env.LANGFUSE_SECRET_KEY,
      publicKey: process.env.LANGFUSE_PUBLIC_KEY,
      baseUrl: process.env.LANGFUSE_HOST,
    }),
  },
});

​
Metrics to Monitor

  • Request Rate: Number of tool calls per minute
  • Latency: Response time for tool execution
  • Error Rate: Failed tool calls and error types
  • Resource Usage: CPU, memory, and network utilization

​
Health Checks

All MCP servers expose a health check endpoint: 
# Health check endpoint
curl https://your-server.run.mcp-use.com/healthz

# Response
{"status": "ok", "timestamp": "2025-02-19T10:30:00Z"}
Use this endpoint for:
  • Load balancer health checks
  • Container orchestration readiness probes
  • Uptime monitoring

​
Next Steps

Deploy to Manufact Cloud

Get started with the fastest deployment option

Self-Host Your Server

Set up your own infrastructure

Docker Deployment

Containerize your MCP server

Integration Guides

Connect to AI assistants

Build docs developers (and LLMs) love

Get started for freeTalk to us