Build MCP Server - Setup Guide

Source: anthropics/claude-plugins-official · 3,148 installs Category: MCP Development / Agent Infrastructure License: MIT · Language: TypeScript, Python · Published: 2026

Anthropic's official skill for building MCP (Model Context Protocol) servers. Step-by-step guidance from transport layer setup through tool registration, resource exposure, and prompt templates. The most-installed MCP server creation guide on skills.sh.

---

What It Does

Phase	What You Build
Transport	stdio or HTTP/SSE transport layer for MCP communication
Tools	Register tools with typed inputs and outputs
Resources	Expose data sources as MCP resources with URI templates
Prompts	Define reusable prompt templates for agent consumption
Testing	Validate server behavior with MCP Inspector

Covers both TypeScript (via @modelcontextprotocol/sdk) and Python (via mcp package) implementations with complete, runnable examples.

---

Prerequisites

• Node.js 18+ (for TypeScript servers) or Python 3.10+ (for Python servers)
• Familiarity with JSON-RPC concepts
• Hermes Agent (or any MCP-compatible client) for testing

---

Installation

Via skills.sh (Recommended)

npx skills add anthropics/claude-plugins-official@build-mcp-server

Direct from GitHub

git clone https://github.com/anthropics/claude-plugins-official.git
cd claude-plugins-official/build-mcp-server

Quick Start - TypeScript

mkdir my-mcp-server && cd my-mcp-server
npm init -y
npm install @modelcontextprotocol/sdk zod
npm install -D typescript @types/node

Quick Start - Python

mkdir my-mcp-server && cd my-mcp-server
python3 -m venv venv
source venv/bin/activate
pip install mcp

---

Usage with Hermes Agent

Once your MCP server is built, register it in Hermes:

# ~/.hermes/config.yaml
mcp_servers:
  my-custom-server:
    command: node
    args: ["dist/index.js"]
    # or for Python:
    # command: python3
    # args: ["server.py"]

Hermes will auto-discover the server's tools and make them available to all agents.

# Verify the server is connected
hermes mcp test
hermes mcp list  # should show your server's tools

---

Core Patterns

Tool Registration (TypeScript)

server.setRequestHandler(ListToolsRequestSchema, async () => ({
  tools: [{
    name: "get_weather",
    description: "Get current weather for a city",
    inputSchema: {
      type: "object",
      properties: {
        city: { type: "string", description: "City name" }
      },
      required: ["city"]
    }
  }]
}));

Tool Registration (Python)

@server.list_tools()
async def list_tools() -> list[Tool]:
    return [
        Tool(
            name="get_weather",
            description="Get current weather for a city",
            inputSchema={
                "type": "object",
                "properties": {
                    "city": {"type": "string"}
                },
                "required": ["city"]
            }
        )
    ]

---

CorpusIQ Use Cases

Use Case	How
Custom data connectors	Build MCP servers for proprietary data sources
Internal tool exposure	Wrap internal APIs as MCP tools for Hermes agents
Multi-agent tool sharing	One MCP server, consumed by all CorpusIQ-connected agents
Client deployments	Build customer-specific MCP servers for white-label

---

Troubleshooting

Issue	Fix
Server not discovered	Verify command/args in config.yaml; check hermes mcp test
Tool calling fails	Validate inputSchema matches what you're sending
Python import errors	Ensure mcp package is in the venv
stdio transport hangs	Use HTTP/SSE transport for long-running servers

---

Companion Skills

• build-mcp-app - Full MCP application development
• build-mcpb - Same guide, Bun runtime
• mcp-developer - Debugging and testing toolkit
• mcp-security-audit - Security scanning for MCP servers

---

← Skills Catalog | Marketplace →

---

Part of the Hermes Skills Library. Curated by CorpusIQ - one MCP endpoint, all your business tools. Content remains attributed to original authors and repositories.

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects