mcp-examples
Provides reference implementations of MCP servers using FastAPI and FastMCP, including calculator and RSS/Atom feed parser tools.
README
mcp-examples
Reference implementations of MCP servers using:
- FastAPI +
fastapi-mcp - FastMCP (STDIO and HTTP transports)
- FastMCP +
feedparserfor RSS/Atom querying
These examples are intentionally small and are useful as starter templates.
Repository layout
.
├── .env-template
├── fastapi-mcp/
│ └── fastapi-mcp-calculator.py
├── fastmcp/
│ ├── fastmcp-calculator.py
│ └── fastmcp-calculator-v2.py
├── mcp-feed-parser/
│ └── mcp-feed-parser.py
└── requirements.txt
Prerequisites
- Python 3.10+
- uv (Python package manager and runner)
- Node.js (optional, only for MCP Inspector)
Environment setup
Create a local environment file from the template:
cp .env-template .env
Then edit .env with your local values.
The provided .env-template includes starter defaults. The current examples do
not require environment variables unless you wire these values into the scripts.
Install dependencies (uv)
uv venv
source .venv/bin/activate
uv pip install -r requirements.txt
Examples
1) FastAPI + FastAPI-MCP calculator
File: fastapi-mcp/fastapi-mcp-calculator.py
What it demonstrates:
- Building a basic FastAPI calculator API (
add,subtract,multiply,divide) - Mounting MCP over the same app using
FastApiMCP(app).mount_http() - Running with Uvicorn on
0.0.0.0:8000
Run:
uv run python fastapi-mcp/fastapi-mcp-calculator.py
2) FastMCP calculator (STDIO transport)
File: fastmcp/fastmcp-calculator.py
What it demonstrates:
- Registering tools with
@mcp.tool() - Naming/tagging tools for better discoverability
- Running an MCP server over STDIO (useful for local client subprocess integration)
Run:
uv run python fastmcp/fastmcp-calculator.py
3) FastMCP calculator (HTTP transport)
File: fastmcp/fastmcp-calculator-v2.py
What it demonstrates:
- The same calculator tools as above
- Running FastMCP with HTTP transport on
localhost:8080
Run:
uv run python fastmcp/fastmcp-calculator-v2.py
4) FastMCP feed parser
File: mcp-feed-parser/mcp-feed-parser.py
What it demonstrates:
- Querying RSS/Atom feeds using
feedparser - Reusing a shared feed-search helper
- Exposing feed lookup as MCP tools:
fcc_news_parserfcc_youtube_parserfcc_secret_message
Run:
uv run python mcp-feed-parser/mcp-feed-parser.py
Inspecting/debugging with MCP Inspector
The MCP Inspector is the easiest way to test tools and inspect schemas.
npx @modelcontextprotocol/inspector <your-server-command>
For example (local Python server):
npx @modelcontextprotocol/inspector uv run python fastmcp/fastmcp-calculator.py
Notes
- The FastAPI sample filename is intentionally kept as-is (
calcularor) to match repository history. - Calculator division tools guard against division by zero.
- The feed parser returns a fallback message when no matches are found.
References:
- MCP documentation and specification: https://modelcontextprotocol.io/
- MCP Inspector guide: https://modelcontextprotocol.io/docs/tools/inspector
@modelcontextprotocol/inspectorpackage: https://www.npmjs.com/package/@modelcontextprotocol/inspector- FastMCP docs: https://gofastmcp.com/getting-started/welcome
- FastMCP repository: https://github.com/prefecthq/fastmcp
- FastAPI-MCP docs: https://fastapi-mcp.tadata.com/
- FastAPI docs: https://fastapi.tiangolo.com/
- feedparser docs: https://feedparser.readthedocs.io/en/latest/
Original reference video
- https://youtu.be/DosHnyq78xY?si=V4DSGMHX4kE3BYMh
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。