free-web-search-ultimate
Universal Search-First Knowledge Acquisition Plugin for LLMs. Enables real-time web search and deep page browsing via MCP or CLI. Zero-cost, privacy-first, supports DuckDuckGo, Bing, Google, Brave, Wikipedia, Arxiv, YouTube, Reddit and more.
README
<div align="center"> <img src="assets/banner.png" alt="Cross-Validated Search" width="100%"/> <h1>🔍 Cross-Validated Search</h1> <p><strong>The Only Search Skill That Prevents LLM Hallucinations</strong></p> <p><em>Multi-source verification. Zero hallucinations. Free forever.</em></p>
<br/>
Stop hallucinations. Start verification.
Every LLM hallucinates facts—Claude, GPT-4, Gemini, Llama—all of them. This plugin introduces Cross-Validated Search: every claim is verified against multiple independent sources before being presented as fact.
🏗️ Architecture
<div align="center"> <img src="assets/architecture.png" alt="Architecture Diagram" width="90%"/> </div>
One plugin, every ecosystem. Whether you use Claude Desktop, Cursor, OpenClaw, or a custom LangChain agent, this plugin connects your LLM to verified facts through cross-validation.
🌟 The Cross-Validation Paradigm
| Old Paradigm (Standard LLM) | New Paradigm (Cross-Validated Search) |
|---|---|
| Answers from training data | Answers from real-time web search |
| May hallucinate facts | Cross-validates facts across sources |
| Single knowledge source | Multiple independent sources |
| No confidence score | Confidence score per fact: ✅ 🟢 🟡 🔴 |
| User must trust blindly | Cites all sources for verification |
When this plugin is installed, the AI agent:
- Never Claims Unverified Facts — Every factual claim is checked against multiple sources
- Cross-Validates — Facts must appear in 2+ sources to be marked as verified
- Assigns Confidence — Each fact gets a confidence score based on source agreement
- Cites All Sources — Every claim comes with verifiable URLs
📦 Installation
pip install cross-validated-search
Requirements: Python 3.10+
Or install from source:
git clone https://github.com/wd041216-bit/cross-validated-search.git
cd cross-validated-search
pip install -e .
🚀 Enhanced Version (with API Providers)
Need more search providers? Check out the with-api-providers branch:
# Install enhanced version with Tavily support
pip install git+https://github.com/wd041216-bit/cross-validated-search@with-api-providers
# Or with Tavily extra
pip install "cross-validated-search[tavily] @ git+https://github.com/wd041216-bit/cross-validated-search@with-api-providers"
Enhanced features:
- 🔍 Tavily API — Premium search quality (requires
TAVILY_API_KEY) - 🔄 Parallel execution — Multiple engines running concurrently
- 📊 Better relevance — Advanced search depth options
Note: The enhanced version requires API keys. The main branch remains free and keyless.
🔌 Integration Guide
Claude Desktop & Cursor (via MCP)
Add to your claude_desktop_config.json or Cursor MCP settings:
{
"mcpServers": {
"cross-validated-search": {
"command": "cross-validated-mcp",
"args": []
}
}
}
OpenClaw (via CLI-Anything)
# Install — the skill is auto-discovered from the bundled SKILL.md
pip install cross-validated-search
LangChain / Custom Agents
from langchain.tools import Tool
import subprocess, json
def cross_validate_search(query: str) -> str:
result = subprocess.run(
["cross-validate", query, "--json"],
capture_output=True, text=True
)
data = json.loads(result.stdout)
return data.get("answer", "No results found.")
search_tool = Tool(
name="cross_validate_search",
func=cross_validate_search,
description="Search the web with cross-validation. Every fact is verified against multiple sources."
)
OpenAI Function Calling
tools = [
{
"type": "function",
"function": {
"name": "cross_validate_search",
"description": "Search the web with multi-source cross-validation. Prevents hallucinations by verifying facts across independent sources.",
"parameters": {
"type": "object",
"properties": {
"query": {"type": "string", "description": "The search query"},
"type": {"type": "string", "enum": ["text", "news", "images", "videos", "books"], "default": "text"}
},
"required": ["query"]
}
}
}
]
💻 CLI Usage
cross-validate — Cross-Validated Web Search
# General knowledge (3+ sources, cross-validated)
cross-validate "What is the population of Tokyo?"
# Breaking news (multiple news sources)
cross-validate "OpenAI GPT-5" --type news --timelimit w
# Images (verified sources)
cross-validate "neural network diagram" --type images
# Chinese search
cross-validate "人工智能最新进展" --region zh-cn
# JSON output for programmatic use
cross-validate "quantum computing" --json
browse-page — Deep Page Reading
# Read full page content
browse-page "https://arxiv.org/abs/2303.08774"
# JSON output
browse-page "https://example.com/article" --json
🎯 Confidence Scoring System
| Score | Meaning | When to Use |
|---|---|---|
| ✅ Verified | 3+ sources agree, high authority | Cite as fact |
| 🟢 Likely True | 2 sources agree, medium confidence | Cite with confidence note |
| 🟡 Uncertain | Single source or minor conflicts | Flag as unverified |
| 🔴 Likely False | Major contradictions or no sources | Do not use |
🏆 Why This Over Alternatives?
| Feature | Cross-Validated Search | Tavily API | Serper API | Bing Search API |
|---|---|---|---|---|
| Cost | Free | $0.01/req | $0.001/req | $3/1000 req |
| Cross-Validation | Yes | No | No | No |
| Confidence Score | Yes | No | No | No |
| Hallucination Prevention | Yes | No | No | No |
| API Key Required | No | Yes | Yes | Yes |
| MCP Support | Yes | Partial | No | No |
| CLI-Anything | Yes | No | No | No |
📄 License
MIT License — free for personal and commercial use.
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。