Tavily is a web search API built specifically for AI agents, and it's noticeably less noisy than throwing raw Google results at your LLM. You get answer synthesis out of the box, domain filtering for when you only want results from .edu or news sites, and depth control via max_results to keep token costs reasonable. The integration is straightforward: set your TAVILY_API_KEY, POST to their endpoint with a query and optional filters, parse the JSON response. Use it when you need current information like news or facts, but skip it if you're working with internal data. The main gotcha is rate limiting on the free tier, so implement backoff logic if you're hammering it in a loop.
npx skills add https://github.com/alphaonedev/openclaw-graph --skill tavily-web-search