MCP Web Tools
Provides local web search and content fetching capabilities for AI assistants, enabling them to search DuckDuckGo and extract clean text from web pages. All requests originate from the user's machine to ensure direct network control and bypass external proxies.
README
MCP Web Tools
A Model Context Protocol (MCP) server that provides web search and fetch capabilities for AI assistants like Claude Code.
This is useful when you want your code-generation tool to make web requests directly from your machine, rather than having those requests proxied through an external server you don't control. The MCP server runs locally on your host, so all web requests originate from your own network.
Features
-
web_search - Search the web using DuckDuckGo
- Returns formatted results with titles, URLs, and snippets
- Configurable result count and region
-
web_fetch - Fetch and extract content from web pages
- Automatic HTML-to-text conversion (removes scripts, styles, navigation)
- JSON response formatting
- Configurable timeout
Requirements
- Python 3.13+
- uv
Installation
Quick Install (Recommended)
# Clone the repository
git clone <repo-url>
cd mcp-web-tools
# Install globally and configure Claude Code
make install
This will:
- Install
mcp-web-toolsas a global executable viauv tool install - Register the MCP server with Claude Code using
claude mcp add - The executable will be available at
~/.local/bin/mcp-web-tools
Restart Claude Code after installation.
Manual Installation
# Install dependencies
uv sync
# Run directly from project
uv run mcp-web-tools
Makefile Targets
| Target | Description |
|---|---|
make install |
Install globally and configure Claude Code |
make uninstall |
Remove the global installation |
make test |
Run tests with pytest |
make clean |
Remove build artifacts |
Claude Code Configuration
After running make install, you can verify the registration:
claude mcp list
The tools will be available in Claude Code as:
mcp__web-tools__web_searchmcp__web-tools__web_fetch
Manual Registration
If you prefer to register manually:
# User-wide (available in all projects)
claude mcp add --scope user web-tools mcp-web-tools
# Project-local (only current project)
claude mcp add --scope project web-tools mcp-web-tools
Alternative: Run from Project Directory
For development, you can register to run directly from the source:
claude mcp add --scope project web-tools uv run --directory /path/to/mcp-web-tools mcp-web-tools
Tool Reference
web_search
Search the web using DuckDuckGo.
Parameters:
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
| query | string | Yes | - | The search query |
| max_results | integer | No | 10 | Maximum number of results |
| region | string | No | "wt-wt" | Region for results (e.g., "us-en", "uk-en") |
Example response:
Search results for: python mcp server
1. Building MCP Servers in Python
URL: https://example.com/article
Learn how to build Model Context Protocol servers...
2. MCP Documentation
URL: https://modelcontextprotocol.io/docs
Official documentation for the Model Context Protocol...
web_fetch
Fetch the content of a web page.
Parameters:
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
| url | string | Yes | - | The URL to fetch |
| timeout | integer | No | 30 | Request timeout in seconds |
Example response:
Content from https://example.com/article:
Building MCP Servers
This guide covers the basics of creating an MCP server...
Development
Running Tests
make test
# Or directly
uv run pytest -v
Project Structure
mcp-web-tools/
src/
mcp_web_tools/
__init__.py # Package metadata
server.py # MCP server implementation
tools.py # Tool implementations (search, fetch)
tests/
test_tools.py # Unit tests
pyproject.toml # Project configuration
Makefile # Install/uninstall automation
Dependencies
- mcp - Model Context Protocol SDK
- ddgs - DuckDuckGo Search API
- httpx - Async HTTP client
- lxml - HTML parsing (optional; falls back to regex-based extraction if not installed)
License
MIT
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。