unsearch

What This Skill Does

The UnSearch skill provides OpenClaw agents with powerful, real-time web search and deep-content extraction capabilities. As an open-source alternative to services like Tavily or Exa, it allows your agent to bridge the gap between static model knowledge and the dynamic, live internet. It enables comprehensive data gathering, fact verification, and autonomous research, making it an essential component for high-quality RAG (Retrieval-Augmented Generation) pipelines and complex knowledge-intensive tasks.

Installation

To integrate UnSearch into your environment, follow these steps:

1. Ensure your OpenClaw environment is configured.
2. Run the following command in your terminal: clawhub install openclaw/skills/skills/rakesh1002/unsearch
3. Obtain your free API key at https://unsearch.dev.
4. Set the environment variable: export UNSEARCH_API_KEY="your_api_key_here".

Use Cases

• Market Intelligence: Automatically track competitors, search news, and scrape product pages to build detailed landscape reports.
• Fact-Checking: Verify claims found in documents by searching multiple, high-authority engines simultaneously.
• Dynamic RAG: Enhance LLM output by fetching live context before answering user queries, reducing hallucination.
• Deep Web Research: Perform iterative, multi-step search queries to compile long-form articles or technical summaries.

Example Prompts

1. "Perform a deep research report on the latest advancements in solid-state battery technology from the last three months, citing at least five different sources."
2. "Search for the documentation of the latest Python version, extract the content from the official release page, and summarize the breaking changes."
3. "Verify if the claim that 'Company X acquired Startup Y' is accurate by searching news outlets and provide the URLs of the confirming articles."

Tips & Limitations

• Optimize for Cost: The service provides 5,000 free queries per month. If you are running high-frequency automated tasks, monitor your usage via the UnSearch dashboard.
• Search Depth: Use the agent/search endpoint with search_depth: 'advanced' when you need higher accuracy and synthesized answers, but stick to basic or fast for quick data retrieval to save time.
• Filtering: Always use include_domains or exclude_domains when possible to limit the noise in your results and improve the signal-to-noise ratio of your scraped content.
• Language Support: While the default is English, ensure you specify the correct ISO language code if you are conducting research in other regions to get geographically relevant results.