Build mcp server

Skill m00kk/agent-skills-playbook/skills/build-mcp-server

15 production Agent Skills — MCP, LangGraph, RAG, security, Cursor SDK. MIT licensed.

Install

npx -y skills add m00kk/agent-skills-playbook --skill build-mcp-server

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Builds production Model Context Protocol (MCP) servers in Python (FastMCP) or TypeScript (@modelcontextprotocol/sdk). Use when the user asks to create an MCP server, expose tools/resources to Cursor or Claude, or integrate APIs via MCP.

SKILL.md

2.3 KB, as published. Nobody here has run it

Build MCP Server

Choose stack

Preference	SDK	Transport
Python / data APIs	`mcp` + FastMCP	stdio (local), streamable HTTP (remote)
Node / TS monorepo	`@modelcontextprotocol/sdk`	same

Default: Python FastMCP unless the project is TypeScript-only.

Workflow

Progress:
- [ ] Define tools (actions) vs resources (read-only context)
- [ ] Scaffold server with schema validation (Pydantic / Zod)
- [ ] Implement handlers with least-privilege defaults
- [ ] Test with MCP Inspector or client config
- [ ] Document install in README (no secrets)

Step 1: Scaffold (Python)

mkdir mcp-my-service && cd mcp-my-service
python -m venv .venv && source .venv/bin/activate
pip install "mcp[cli]" pydantic

Minimal server pattern:

from mcp.server.fastmcp import FastMCP

mcp = FastMCP("my-service")

@mcp.tool()
def search_items(query: str, limit: int = 10) -> str:
    """Search items by query. Returns JSON string."""
    # implement; return json.dumps(results)
    ...

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

Step 2: Tool design rules

One tool = one user-visible action; name verbs (search_items, not helper)
Input schemas strict; reject unknown fields
Return structured text (JSON string) for agent parsing
Never return raw secrets or full env dumps

Step 3: Client config (Cursor)

Add to MCP settings (project or user):

{
  "mcpServers": {
    "my-service": {
      "command": "python",
      "args": ["-m", "mcp_my_service"],
      "cwd": "${workspaceFolder}"
    }
  }
}

Step 4: Verify

Server starts without import errors
Each tool callable with valid + invalid inputs (invalid must fail clearly)
Remote deploy: OAuth, rate limits, health check, no secrets in tool responses

Anti-patterns

Dumping entire databases via one giant query tool
Shell execution without argument allowlists
Hardcoded API keys (use env vars documented in .env.example only)