返回 MCP 目录
public公开dns本地运行

MCP模板

一个简单的模板,用于使用模型上下文协议为 Cursor IDE 创建自定义工具,并可通过 Heroku、Docker 或直接在 Cursor IDE 内部署。

article

README

MCP Server Template for Cursor IDE

A simple template for creating custom tools for Cursor IDE using Model Context Protocol (MCP). Create your own repository from this template, modify the tools, and connect them to your Cursor IDE.

Server Mood Response

Quick Start

  1. Click "Deploy to Heroku" button

    Deploy to Heroku

  2. After deployment, configure Cursor:

    • Open Cursor Settings → Features
    • Add new MCP server
    • Use your Heroku URL with /sse path (e.g., https://<your-app-name>.herokuapp.com/sse)
  3. Test your agent's mood in Cursor:

    • Ask your agent "Please ask about our server mood and let me know how it is."
    • The server will respond with a cheerful message and a heart ❤️

Alternative Setup Methods

You can run the server in three ways: using Docker, traditional Python setup, or directly in Cursor IDE.

Docker Setup

The project includes Docker support for easy deployment:

  1. Initial setup:
# Clone the repository
git clone https://github.com/kirill-markin/weaviate-mcp-server.git
cd weaviate-mcp-server

# Create environment file
cp .env.example .env
  1. Build and run using Docker Compose:
# Build and start the server
docker compose up --build -d

# View logs
docker compose logs -f

# Check server status
docker compose ps

# Stop the server
docker compose down
  1. The server will be available at:

    • SSE endpoint: http://localhost:8000/sse
  2. Quick test:

# Test the server endpoint
curl -i http://localhost:8000/sse
  1. Connect to Cursor IDE:
    • Open Cursor Settings → Features
    • Add new MCP server
    • Type: Select "sse"
    • URL: Enter http://localhost:8000/sse

Traditional Setup

First, install the uv package manager:

# Install uv on macOS
brew install uv
# Or install via pip (any OS)
pip install uv

Start the server using either stdio (default) or SSE transport:

# Install the package with development dependencies
uv pip install -e ".[dev]"

# Using stdio transport (default)
uv run mcp-simple-tool

# Using SSE transport on custom port
uv run mcp-simple-tool --transport sse --port 8000

# Run tests
uv run pytest -v

After installation, you can connect the server directly to Cursor IDE:

  1. Right-click on the cursor-run-mcp-server.sh file in Cursor
  2. Select "Copy Path" to copy the absolute path
  3. Open Cursor Settings (gear icon)
  4. Navigate to Features tab
  5. Scroll down to "MCP Servers"
  6. Click "Add new MCP server"
  7. Fill in the form:
    • Name: Choose any name (e.g., "my-mcp-server-1")
    • Type: Select "stdio" (not "sse" because we run the server locally)
    • Command: Paste the absolute path to cursor-run-mcp-server.sh that you copied earlier. For example: /Users/kirillmarkin/weaviate-mcp-server/cursor-run-mcp-server.sh

Environment Variables

Available environment variables (can be set in .env):

  • MCP_SERVER_PORT (default: 8000) - Port to run the server on
  • MCP_SERVER_HOST (default: 0.0.0.0) - Host to bind the server to
  • DEBUG (default: false) - Enable debug mode
  • MCP_USER_AGENT - Custom User-Agent for website fetching

Additional options

Installing via Smithery

smithery badge

To install MCP Server Template for Cursor IDE for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @kirill-markin/example-mcp-server --client claude

Glama server review

Server Template for Cursor IDE MCP server

help

运行方式说明

cloud

托管运行

托管运行通常表示这个 MCP Server 由服务方环境承载,用户一般按页面提供的连接方式或授权流程接入,不需要在本地长期启动一个 MCP 进程

  1. 打开服务方连接页
  2. 完成授权或复制端点
  3. 在 MCP 客户端中连接
terminal

本地运行 / 其它方式

本地运行通常需要用户在自己的电脑或服务器上安装依赖,把 server_config 复制到 MCP 客户端,并按 env_schema 补齐环境变量、密钥或其它配置

  1. 复制 server_config
  2. 安装所需依赖
  3. 补齐环境变量后重启客户端