用于匿名搜索和浏览网页的工具
概览
用于你的大型语言模型：世界上第一个模型上下文协议（MCP）服务器，提供通过Tor匿名搜索和获取网页的工具。搜索结果来自DuckDuckGo，页面内容通过实际的Tor浏览器实例检索，保持了通用指纹，使Tor用户彼此难以区分。 数百万人通过LM Studio或Ollama等应用使用本地LLM。本地运行模型提供了极其私密且低成本的方式，用于获取信息、探索想法和自动化计算。 https://github.com/user-attachments/assets/1e07d780-8f9e-474e-a7f0-7581c4d64fcb 搜索工具在大型语言模型聊天界面中变得无处不在——ChatGPT、Gemini等大多数查询都会自动在后台使用它们。
概述（AI 翻译）
为你的大型语言模型提供绝对匿名知识检索：全球首个模型上下文协议（MCP）服务器，提供通过Tor匿名搜索和获取网页的工具。搜索结果来自DuckDuckGo，页面内容通过真实的Tor浏览器实例检索，保持了通用指纹，使Tor用户彼此难以区分。
为什么会有这样的存在
数百万人通过LM Studio或Ollama等应用使用本地LLM。本地运行模型提供了极其私密且低成本的方式，用于获取信息、探索想法和自动化计算。
https://github.com/user-attachments/assets/1e07d780-8f9e-474e-a7f0-7581c4d64fcb
搜索工具在大型语言模型聊天界面中变得无处不在——ChatGPT、Gemini等大多数查询都会自动在后台使用。与此同时，所有推荐用于本地LLM的网络搜索集成工具，都通过现成的Google Chrome、Firefox或Brave安装请求，这些工具（尽管宣传上如此）会留下特定的浏览器指纹，可以用来追踪、监控、限制速率或地理限制。
本页由 AI 翻译，技术细节请以原文为准。
README
以下为原文 README 节选，完整内容请查看原文。
Onion Search MCP
Absolutely anonymous knowledge retrieval for your LLM: the world’s first Model Context Protocol (MCP) server that exposes tools to search the web and fetch pages anonymously through Tor. Search results come from DuckDuckGo, and page content is retrieved through an actual Tor Browser instance, preserving the universal fingerprint that makes Tor users indistinguishable from one another.
Why this exists
Millions of people use local LLMs through apps like LM Studio or Ollama. Running models locally offers an extremely private and low-cost way to access information, explore ideas, and automate computation.
https://github.com/user-attachments/assets/1e07d780-8f9e-474e-a7f0-7581c4d64fcb
Search tools are becoming ubiquitous in LLM chat interfaces—ChatGPT, Gemini, and others use them automatically in the background for most queries. Meanwhile, every web search integration tool recommended for local LLMs route requests through off-the-shelf installations of Google Chrome, Firefox, or Brave, which (despite their marketing) leave specific browser fingerprints that can be used to track, surveil, rate-limit, or geo-restrict you.
This tool takes a different approach: every request flows through the Tor network and browser, which routes your requests through an anonymity network and makes your traffic indistinguishable from millions of other users.
Installation
⚡ Install quickly by giving this prompt to Claude Code, Codex, or any other agent of choice.
Prerequisites
Download Python 3.11+. Check if you have it installed by running in your terminal:
python3 --versionon macOS or Linuxpython --versionon Windows
If you don’t, download and install the latest version from python.org.
On Linux, link-copy mode in the TUI (C key) requires xclip:
- Debian/Ubuntu:
sudo apt install xclip - Fedora:
sudo dnf install xclip - Arch:
sudo pacman -S xclip
Download the project
For those comfortable with Git, clone the repository:
git clone https://github.com/maximilianromer/Onion-Search-MCP.git
For the less technical: download the project by clicking the green “Code” button above and selecting “Download ZIP”; then extract the ZIP file to a location to keep it in.
⚠️ Warning: To speed up initialization, Tor network information is cached in this folder. I highly discourage saving this MCP in a directory that backs up to cloud storage.
Run the installer
安装
This server does not publish a one-line install command.
Open the repository installation guide 新手？先看 mcpskill.org 教程 →配置
{
"mcpServers": {
"onion-search-mcp": {
"command": "/Users/maxro/Documents/Onion-Search-MCP/.venv/bin/python",
"args": [
"/Users/maxro/Documents/Onion-Search-MCP/server.py"
]
}
}
}