MCPFier transforms any command, script, or tool into a standardized MCP (Model Context Protocol) server that LLMs can use seamlessly.

A complete MCP-to-API gateway supporting three execution modes: local commands, containerized execution, and webhook/API calls. Ships with embedded analytics, authentication, and a web dashboard.

Think "GitHub Actions for MCP" with enterprise features - configure once, use everywhere with comprehensive monitoring and security.

1. Configure commands in config.yaml:

   commands:
     - name: get-weather
       script: curl
       args: ["https://wttr.in/?format=3"]
       description: "Get current weather"

2. Generate setup instructions:

   ./mcpfier --setup

3. Add to Claude Desktop using the JSON configuration from step 2

4. Use in Claude: "Use the get-weather tool"

• Universal Tool Interface: Transform any script into an MCP tool
• Three Execution Modes: Local commands, Docker containers, and HTTP webhooks/APIs
• Dual Transport: STDIO for desktop, HTTP/SSE for enterprise deployments
• Complete MCP-to-API Gateway: Full upstream API integration with authentication
• Secret Management: Environment variable expansion for secure credential handling
• Rate Limiting: Token bucket rate limiting per client (API key or IP)
• Authentication Ready: API keys with granular per-tool permissions
• Embedded Analytics: SQLite-based analytics with real-time web dashboard
• Enterprise Ready: Multi-client support, request logging, monitoring
• MCP Compliant: Based on mcp-go v0.42.0

Commands are defined in config.yaml:

commands:
  # Local execution
  - name: list-files
    script: ls
    args: ["-la"]
    description: "List directory contents"
    timeout: "10s"

  # Containerized execution  
  - name: python-analysis
    script: python
    args: ["/app/analyze.py"]
    description: "Run data analysis"
    container: "python:3.9-slim"
    timeout: "5m"
    env:
      DATA_SOURCE: "production"

  # Webhook/API execution
  - name: httpbin-get
    description: "Test GET request to httpbin.org"
    timeout: "10s"
    webhook:
      url: "https://httpbin.org/get"
      method: "GET"
      headers:
        User-Agent: "MCPFier/1.0"

# Server configuration (HTTP mode)
server:
  http:
    enabled: true
    host: "localhost"
    port: 8080
    auth:
      enabled: true
      mode: "simple"
      api_keys:
        "mcpfier_dev_123456":
          name: "Development Key"
          permissions: ["*"]

# Analytics configuration
analytics:
  enabled: true
  database_path: "~/.mcpfier/analytics.db"
  retention_days: 30

*Either script or webhook must be specified.

Rate limiting is applied per client, identified by API key or IP address.

Webhook authentication supports environment variable expansion for secure credential handling:

commands:
  - name: github-api
    webhook:
      url: "https://api.github.com/user"
      auth:
        type: "bearer"
        token: "${GITHUB_TOKEN}"  # Reads from environment

Set the environment variable before starting MCPFier:

export GITHUB_TOKEN="ghp_your_token_here"
./mcpfier --server

Supported in all webhook auth fields: token, key, user, pass.

MCPFier searches for config.yaml in the following order:

1. Path specified by MCPFIER_CONFIG environment variable
2. Current working directory (./config.yaml)
3. Directory containing the mcpfier executable
4. User's home directory (~/.mcpfier/config.yaml or ~/mcpfier/config.yaml)
5. System-wide location (/etc/mcpfier/config.yaml)

Use --config <path> to specify a custom configuration file path.

./mcpfier [OPTIONS] [COMMAND]

Options:
  --config, -c <path>    Specify custom configuration file path
  --mcp                  Start MCP local STDIO mode
  --server               Start MCP server mode (HTTP)
  --setup                Generate Claude Desktop configuration
  --analytics            Show usage analytics and statistics
  --help, -h             Show help information


Examples:
  ./mcpfier --config ./my-config.yaml --mcp
  ./mcpfier -c /etc/mcpfier/config.yaml echo-test
  ./mcpfier --analytics --config ~/.mcpfier/config.yaml

git clone https://github.com/gleicon/mcpfier.git
cd mcpfier
go build -o mcpfier

# Build image
docker build -t mcpfier:latest .

# Run with persistent volumes
docker run -d --name mcpfier -p 8080:8080 \
  -v /opt/mcpfier/config:/app/config:ro \
  -v /opt/mcpfier/data:/app/data \
  mcpfier:latest

For complete Docker deployment guide, see DOCKER.md

MCPFier supports two transport modes to fit different use cases:

Perfect for local development and Claude Desktop integration:

./mcpfier --mcp                    # Start STDIO MCP server (default)
./mcpfier --setup                  # Generate Claude Desktop config
./mcpfier --mcp --config /path    # Use custom config file

For production deployments, multiple clients, and enterprise features:

./mcpfier --server                 # Start HTTP MCP server with authentication
./mcpfier --server --config /path # HTTP server with custom config

For complete HTTP server setup and authentication guide, see SERVER.md

./mcpfier command-name             # Execute command directly
./mcpfier --config /path cmd-name  # Use custom config for command

View usage statistics and command performance:

./mcpfier --analytics               # Show analytics for last 7 days
./mcpfier --analytics --config /path # Use custom config for analytics

Analytics tracks:

• Command execution statistics (local, container, webhook modes)
• HTTP server metrics with request/response tracking
• Authentication success and failure rates
• Upstream API call success rates and latencies
• Error categorization and breakdown by type
• Real-time performance monitoring

The embedded web dashboard provides comprehensive analytics when running in HTTP server mode. Access at http://localhost:8080/mcpfier/analytics (default configuration).

Note: The analytics database directory must exist before starting MCPFier. Create the directory manually:

mkdir -p ~/.mcpfier  # For default config

MCPFier implements a modular architecture with three execution engines:

• Configuration: YAML-based command definitions with auto-discovery
• Transport: Dual-mode support (STDIO for desktop, HTTP for enterprise)
• Execution Engines: Local commands, Docker containers, and HTTP webhooks/APIs
• Analytics Engine: SQLite-based embedded analytics with web dashboard
• Authentication: API key-based authentication with granular permissions
• MCP Server: MCP 2025-06-18 compliant
• Security: Container isolation, authentication, request validation, resource limits

• Local Mode: Fast execution, same privileges as MCPFier
• Container Mode: Complete isolation, no host access, resource limits
• Best Practices: Use containers for untrusted code, define timeouts, limit privileges

• Enterprise Integration: Transform internal APIs and workflows into MCP tools
• Development Tools: Linting, testing, building, deployment automation
• Infrastructure: Health checks, log analysis, backup operations
• API Gateway: Bridge MCP clients to existing REST APIs and webhooks
• Hybrid Workflows: Combine local scripts, containerized tools, and external APIs
• AI Agents: Enable LLMs to access enterprise systems and external services

go test ./...              # Run all tests
./mcpfier echo-test        # Test CLI mode
./mcpfier --setup          # Test MCP configuration
./mcpfier --analytics      # Test analytics (requires ~/.mcpfier/ directory)

• SERVER.md - HTTP server setup and authentication guide
• DOCKER.md - Container deployment with persistent volumes
• SECURITY.md - Security best practices and configuration
• SETUP.md - Detailed setup instructions
• ARCHITECTURE.md - Technical architecture
• AUTHENTICATION.md - Technical details on authentication

• MCP Specification - Official Model Context Protocol 2025-06-18 specification
• mcp-go Library - Go library for MCP implementation used by MCPFier
• Claude Desktop - Desktop client that works with MCPFier STDIO mode

We welcome contributions! Please see our documentation for:

• Architecture decisions and design patterns
• Security considerations and testing requirements
• Roadmap and planned features

MIT License - see LICENSE file for details