tavily-search

What This Skill Does

The tavily-search skill integrates the Tavily AI search engine directly into the OpenClaw agent workspace. Unlike generic web scrapers, Tavily is purpose-built for AI agents, providing clean, indexed content that is ready for LLM consumption. It acts as a powerful alternative to standard web_search tools, offering highly relevant search results complete with titles, URLs, and concise snippets. The skill supports multiple output formats, including raw JSON for programmatic handling, a stable 'brave' schema for compatibility with existing workflows, and a human-readable Markdown format for direct user interaction.

Installation

To integrate this skill, ensure you have the OpenClaw environment set up. Run the installation command: clawhub install openclaw/skills/skills/jacky1n7/openclaw-tavily-search. Once installed, you must provide your API credentials to enable the service. Create or edit the ~/.openclaw/.env file and add your key: TAVILY_API_KEY=your_api_key_here. Alternatively, you can set the TAVILY_API_KEY environment variable directly in your current terminal session. Without these credentials, the script will be unable to authenticate with the Tavily API.

Use Cases

This skill is indispensable for research-heavy tasks. Use it when you need to verify current events, perform background research on technical topics, or find specific documentation that is not included in the agent's initial knowledge base. It is particularly effective for generating synthesis reports where multiple sources are required to build a balanced answer. Because it returns condensed snippets, it is an efficient tool for rapidly gathering context without the overhead of processing massive raw HTML files from broad web crawls.

Example Prompts

1. "Search for the latest updates on the OpenClaw roadmap and summarize the top three features announced."
2. "Look up the recent documentation for Python's latest stable release and provide me with the official URL and a brief summary."
3. "Find sources on the benefits of asynchronous programming in Python and list them as a markdown summary."

Tips & Limitations

To maintain high performance and minimize latency, keep max-results between 3 and 5. Requesting too many results can lead to excessive token usage and slower response times. If you are building automated pipelines, use the --format brave flag to ensure your data schema remains consistent with other OpenClaw search tools. Be aware that this skill relies on an external API; if the API is down or your limit is reached, the skill will return an error. Always verify critical information against primary sources if the search results appear ambiguous or lack sufficient context.