Overview

Tavily Search provides AI-optimized web search via the Tavily API, returning concise, relevant results designed for AI agents. It speeds up research and decision-making by surfacing clean snippets from the web. Note: requires a Tavily API key (TAVILY_API_KEY).

How This Skill Works

The tool queries the Tavily API through Node.js scripts (search.mjs) using your query and optional flags (-n, --deep, --topic, --days). For deeper research, use --deep; for current events, use --topic news and optionally --days. You can also extract full content from a URL with extract.mjs to fetch richer text beyond snippets.

When to Use It

• When you need quick, concise web results to inform an AI agent's reasoning.
• When tackling complex questions and you want deeper, more comprehensive results (--deep).
• When researching current events using --topic news and optionally --days to limit recency.
• When you want to limit results to a specific number with -n for faster processing.
• When you need to fetch full content from a URL using extract.mjs for deeper analysis.

Quick Start

1. Step 1: Run node {baseDir}/scripts/search.mjs "your query"
2. Step 2: Tweak results with flags like -n 10, --deep, --topic news, or --days 7
3. Step 3: (Optional) Run node {baseDir}/scripts/extract.mjs "https://example.com/article" to fetch full content

Best Practices

• Use --deep for complex research questions to obtain more thorough results.
• For current-events work, combine --topic news with --days to keep results fresh.
• Always set the TAVILY_API_KEY in your environment before running searches.
• Start with a small -n (e.g., 5) to prototype, then raise up to the max 20 if needed.
• If you need full article text, supplement snippets with extract.mjs on the target URL.

Example Use Cases

• AI agent searches for 'image generation safety guidelines' with -n 8 and --deep to gather diverse, concise sources.
• Researcher runs a general inquiry with -n 5 to collect quick snippets on 'privacy by design'.
• Journalist uses node scripts with --topic news and --days 7 to fetch recent articles for a developing AI policy story.
• QA bot fetches 'latency benchmarks' with -n 6 to assemble a concise comparison of sources.
• Data scientist extracts content from https://example.com/article using extract.mjs to feed a deeper text analysis pipeline.

Frequently Asked Questions