MCP Servers

A monorepo of Model Context Protocol servers that extend AI assistants with real-world capabilities.

Architecture

graph TB
    subgraph "AI Assistant"
        Client["MCP Client(Claude, etc.)"]
    end

    subgraph "MCP Servers"
        FS["filesystem-mcpFile Operations"]
        GH["github-mcpGitHub API"]
        HTTP["http-client-mcpHTTP Requests"]
        PG["postgres-mcpPostgreSQL"]
    end

    subgraph "External"
        FS_DISK["Local Filesystem"]
        GH_API["GitHub API"]
        HTTP_API["External APIs"]
        PG_DB["PostgreSQL"]
    end

    Client -- "stdio / SSE" --> FS
    Client -- "stdio / SSE" --> GH
    Client -- "stdio / SSE" --> HTTP
    Client -- "stdio / SSE" --> PG

    FS --> FS_DISK
    GH --> GH_API
    HTTP --> HTTP_API
    PG --> PG_DB

Each server communicates with AI assistants via the MCP protocol using stdio (default) or SSE (HTTP) transport.

Servers

Quick Start

# Install dependencies for a specific server
cd filesystem-mcp && npm install

# Development mode (stdio)
npm run dev

# Build for production
npm run build && npm start

Environment

Each server uses a .env file (copy from .env.example) with its own configuration. Common variables:

Tech Stack

• Runtime: Node.js 18+
• Language: TypeScript
• Framework: @modelcontextprotocol/sdk
• Validation: Zod
• Testing: Vitest
• Formatting: Prettier
• Linting: ESLint

Docker

Each server includes a Dockerfile and docker-compose.yml:

cd filesystem-mcp && docker compose up --build