Agent Skills: MCP Server Development

MCP Server Development

Build MCP (Model Context Protocol) servers with Python FastMCP and TypeScript SDK. Covers tools, resources, prompts, transport, and security.

Python FastMCP

Basic Server

from fastmcp import FastMCP

mcp = FastMCP("my-server")

@mcp.tool()
def calculate(expression: str) -> float:
    """Evaluate a mathematical expression."""
    return eval(expression)

@mcp.resource("config://settings")
def get_settings() -> str:
    """Return server configuration."""
    return '{"version": "1.0"}'

if __name__ == "__main__":
    mcp.run()

TypeScript SDK

Basic Server

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";

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

server.tool("calculate", { expression: z.string() }, async ({ expression }) => ({
  content: [{ type: "text", text: String(eval(expression)) }]
}));

const transport = new StdioServerTransport();
await server.connect(transport);

Tool Registration

@mcp.tool()
def search_docs(query: str, limit: int = 5) -> list[dict]:
    """Search documentation for relevant information."""
    return vector_store.search(query, top_k=limit)

Input Validation

from pydantic import BaseModel, Field

class SearchParams(BaseModel):
    query: str = Field(..., min_length=1, max_length=500)
    limit: int = Field(default=5, ge=1, le=20)

Resource Exposure

@mcp.resource("docs://{path}")
def read_doc(path: str) -> str:
    """Read a documentation file."""
    return Path(f"docs/{path}.md").read_text()

Prompt Templates

@mcp.prompt()
def analyze_code(file_path: str, issue: str) -> str:
    """Generate a prompt for code analysis."""
    return f"Analyze {file_path} for: {issue}"

Transport Configuration

| Transport | Use Case | Setup | |-----------|----------|-------| | stdio | Local CLI tools | Default | | HTTP | Web servers | mcp.run(transport="http") | | SSE | Real-time streaming | mcp.run(transport="sse") |

# HTTP transport
mcp.run(transport="http", host="0.0.0.0", port=8000)

# SSE transport
mcp.run(transport="sse", host="0.0.0.0", port=8000)

Error Handling

from fastmcp import ToolError

@mcp.tool()
def risky_operation(data: str) -> str:
    try:
        result = process(data)
        return result
    except ValueError as e:
        raise ToolError(f"Invalid input: {e}")
    except Exception as e:
        raise ToolError(f"Internal error: {e}")

Security Best Practices

• Validate all inputs with Pydantic/Zod
• Never expose secrets in resources
• Rate limit tool calls if public
• Sandbox file system access
• Log all tool calls for audit
• Use HTTPS for HTTP/SSE transports

Testing

def test_calculate_tool():
    result = mcp.tools["calculate"].execute(expression="2 + 2")
    assert result == 4.0

def test_invalid_input():
    with pytest.raises(ToolError):
        mcp.tools["calculate"].execute(expression="invalid")