mcp-builder

Files (6)

• QUICK_REFERENCE.md
• README.md
• SKILL.md
• nodejs_mcp_server.md · reference/
• python_mcp_server.md · reference/
• tool_design_patterns.md · reference/

SKILL.md (first lines)

# MCP Builder

Build production-ready Model Context Protocol (MCP) servers with comprehensive planning, implementation, testing, and deployment guidance.

## Process

### 🚀 High-Level Workflow

1. **Phase 1: Deep Research and Planning** — Understand requirements, study MCP architecture, plan server structure
2. **Phase 2: Implementation** — Write core server code, implement tools/resources, add error handling
3. **Phase 3: Validation** — Test locally, verify tool outputs, debug issues
4. **Phase 4: Deployment** — Containerize (if needed), set environment variables, deploy to production

### Phase 1: Deep Research and Planning

#### 1.1 Clarify Requirements

- What tools/resources should the server expose?
- What is the primary use case?
- Are there authentication or permission requirements?
- What are the scale/performance expectations?

#### 1.2 Choose SDK and Framework

- **Python**: Use the official Python SDK for MCP
  - Reference: [Python SDK README](https://raw.githubusercontent.com/modelcontextprotocol/python-sdk/main/README.md)
  - Best for: Data processing, AI integrations, scientific computing
  - Framework: FastMCP (simplified decorator-based approach) or standard SDK

- **Node.js/TypeScript**: Use the official Node.js SDK
  - Best for: Web APIs, real-time features, JavaScript ecosystem integration
  - Framework: Full SDK with fine-grained control

- **Other languages**: Check [MCP Implementations](https://github.com/modelcontextprotocol#implementations) for community SDKs

#### 1.3 Study Framework Documentation

- **Python SDK**: Understand `Server` class, `@tool` decorator, `@resource` decorator, error handling patterns
- **Node.js SDK**: Understand server initialization, `Tool` and `Resource` classes, request/response handling
- Review reference implementation guides: See `reference/python_mcp_server.md` or `reference/nodejs_mcp_server.md`

#### 1.4 Plan Server Structure

```
my-mcp-server/
├── server.py (or server.ts for Node.js)      # Main server entry point
├── tools/
│   ├── __init__.py
│   ├── tool_1.py                              # Each tool in its own module
│   ├── tool_2.py
│   └── shared_utils.py                        # Shared utility functions
├── resources/
│   ├── __init__.py
│   ├── resource_1.py                          # Each resource in its own module
│   └── templates.py                           # Template definitions
├── requirements.txt (Python) or package.json  # Dependencies
├── .env.example                               # Environment variable template
├── README.md                                  # Usage documentation
└── Dockerfile (optional)                      # For containerization
```

### Phase 2: Implementation

#### 2.1 Create Server Instance

Initialize the MCP server with proper configuration.

**Python (FastMCP - Recommended):**

```python
from mcp.server.fastmcp import FastMCP

server = FastMCP("server-name", instructions="Brief description of what this server does")
```

**Python (Standard SDK):**

```python
from mcp.server import Server
from mcp.types import Tool, Resource

server = Server("server-name")
```

**Node.js:**

```typescript
import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";

const server = new Server({
  name: "server-name",
  version: "1.0.0",
});
```

#### 2.2 Implement Tools

A tool is a callable function that Claude can invoke. Always include:

- Clear, descriptive name
- Detailed description of what it does
- Input schema with types and documentation
- Error handling
- Return value documentation

**Python:**

```python
@server.tool()
def tool_name(arg1: str, arg2: int) -> str:
    """
    Brief description of what this tool does.

    Args:
        arg1: Description of arg1
        arg2: Description of arg2

    Returns:

From baralmanish/AI-Agent-Skills, Apache-2.0.

More from baralmanish/AI-Agent-Skills