A versatile web search and deep research skill for AI agents providing real-time data, full-page scraping, and fact-checking capabilities.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install unsearch
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install unsearch using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is UnSearch Web Search?
UnSearch is a high-performance web search and content extraction tool specifically engineered for AI agents, RAG pipelines, and LLM-driven applications. Serving as a robust, open-source alternative to services like Tavily or Exa, it allows developers to integrate comprehensive web intelligence into their Openclaw Skills projects. By aggregating data from major search engines like Google, Bing, and DuckDuckGo, UnSearch provides a unified interface for real-time information retrieval.
The skill goes beyond basic searching by offering automated content scraping, AI-synthesized research reports, and sophisticated fact verification. Whether you are building an autonomous researcher or a chatbot that needs to verify claims against live data, UnSearch provides the depth and accuracy required for modern AI workflows. Its design ensures that agents can move beyond static training data and interact with the most current information available on the web.
UnSearch Web Search Use Cases
- Real-time Information Retrieval: Access current news, technical documentation, and web data for RAG-based applications.
- Autonomous Research: Conduct deep, multi-source investigations with AI-synthesized summaries for complex topics.
- Fact Verification: Automatically verify claims against multiple web sources to ensure accuracy in LLM outputs.
- Web Scraping: Extract clean, full-page text from specific URLs for deep data analysis or processing.
- AI Agent Enhancement: Provide autonomous agents with the ability to browse the web and answer questions based on real-time search results.
How UnSearch Web Search Works
- Query Processing: The AI agent or application sends a search query to the UnSearch API via dedicated search or agent-search endpoints.
- Multi-Source Aggregation: UnSearch queries multiple search engines simultaneously to find the most relevant and diverse results.
- Content Extraction: The skill optionally scrapes full-page content from the top results, removing noise to provide clean text for AI processing.
- Synthesis and Analysis: For research and verification tasks, the system uses internal AI models to analyze the gathered data and generate structured summaries or verdicts.
- Structured Response: The final data is delivered as a formatted JSON object, ready for the agent to ingest and use in its reasoning chain.
UnSearch Web Search Setup
To integrate UnSearch into your Openclaw Skills environment, you first need an API key from the official dashboard. Once obtained, configure your environment variables as follows:
# Set your UnSearch API Key
export UNSEARCH_API_KEY="uns_your_api_key"
You can test your setup by running a simple query using curl to ensure the connection is established:
curl -X POST "https://api.unsearch.dev/api/v1/search" \
-H "X-API-Key: $UNSEARCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{"query": "latest Openclaw Skills updates", "max_results": 5}'
UnSearch Web Search Data Schema & Taxonomy
UnSearch uses a highly structured data schema to ensure compatibility with various AI agent architectures. Key response objects include:
| Component | Description |
|---|---|
| Results Array | Contains title, URL, snippet, and optional scraped_content for each web result. |
| Agent Answer | A generated AI summary included when include_answer is enabled in agent-search. |
| Research Metadata | Includes executive_summary, key_findings, and a list of verified sources for deep queries. |
| Verification Verdict | Categorizes claims as true, false, partially_true, or misleading with a confidence score. |
All responses include processing_time_ms to help developers optimize their Openclaw Skills implementations.
UnSearch Web Search Advanced Features
- Privacy-First Mode: Enable zero-retention for sensitive searches using the
X-Zero-Retentionheader. - Tavily Compatibility: Drop-in replacement for existing agent workflows via the
/agent/searchendpoint. - Configurable Research Depth: Choose from
quick,standard,deep, orcomprehensivelevels to balance latency and thoroughness. - Fact-Checking Engine: Specialized endpoint for verifying claims with supporting evidence and confidence scores.
- Self-Hosting Support: Deploy the entire UnSearch infrastructure locally for complete data sovereignty and specialized use cases.
SKILL.md
Loading
Related Openclaw Skills
Go2.gg Edge-Native URL Shortener
A high-performance link management tool for shortening URLs, tracking real-time analytics, and generating customizable QR codes at the edge.
AudioPod AI
A professional-grade AI audio processing suite for music generation, vocal isolation, transcription, and text-to-speech.
Tavily Search
An AI-optimized search engine skill that provides clean, LLM-ready content and deep research capabilities for agentic workflows.
Daily Digest
A comprehensive automation skill that aggregates your email, calendar, and daily news into a single, structured report.
last30days
A real-time research tool that extracts data from X, YouTube, and the web to generate AI-synthesized briefings and prompts.
Cannabis Cultivator
A professional meta-search assistant designed to find, filter, and compare cannabis seeds across reputable European seedbanks.
Featured*
