# Tavily MCP Search

> Search the web with LLM-optimized results through the local Tavily MCP server. Use this skill when the user wants to search the web, find articles, look up current information, get recent news, discover sources, or says "search for", "find me", "look up", "what's the latest on", or "find articles about". Returns relevant results with snippets, scores, metadata, and optional raw content or images.

- Skill: `minhngoc25a/tavily-mcp-search` (Agent Skill)
- Install (CLI): `npx skillmds@latest add minhngoc25a/tavily-mcp-search`
- Raw SKILL.md: https://api.skillmd.com/api/skills/minhngoc25a/tavily-mcp-search/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: minhngoc25a (https://skillmd.com/u/minhngoc25a)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/minhngoc25a/tavily-mcp-search

---


# Tavily MCP search

Use `mcp__tavily-mcp__tavily_search` for web search when you do not already have a specific URL.

## When to use

- The user asks for current or external web information.
- You need to discover pages before extracting content.
- You need recent results, source candidates, or domain-filtered search results.

For specific known URLs, use `tavily-mcp-extract` instead. For deep synthesis, use `tavily-mcp-research` instead.

## Parameters

| Parameter | Use |
| --- | --- |
| `query` | Search query. Keep it concise and under 400 characters when possible. |
| `search_depth` | `ultra-fast`, `fast`, `basic`, or `advanced`. Use `basic` by default and `advanced` for higher precision. |
| `max_results` | Number of results, 5-20. Use fewer results for quick lookups. |
| `topic` | Local MCP search supports `general`. |
| `time_range` | `day`, `week`, `month`, or `year` for recency. |
| `start_date`, `end_date` | Date bounds in `YYYY-MM-DD` format. |
| `include_domains` | Restrict results to trusted domains. |
| `exclude_domains` | Exclude domains the user does not want. |
| `country` | Boost results from a named country. |
| `exact_match` | Require exact quoted phrases from the query. |
| `include_raw_content` | Include cleaned page content in each result when snippets are insufficient. |
| `include_images` | Include image results. |
| `include_image_descriptions` | Include AI-generated descriptions for image results. |
| `include_favicon` | Include favicon URLs. |

## Search depth selection

| Depth | Best for |
| --- | --- |
| `ultra-fast` | Low-latency checks where broad relevance is acceptable. |
| `fast` | Fast searches with good relevance. |
| `basic` | General-purpose search. |
| `advanced` | Specific facts, high precision, or important research leads. |

## Patterns

### Quick lookup

Use `basic` depth and 5 results.

### Recent news or updates

Set `time_range` to `day`, `week`, `month`, or `year`.

### Trusted-source search

Use `include_domains` when the user names source constraints or when authoritative sources matter.

### Search then read

Use search to identify promising URLs, then use `mcp__tavily-mcp__tavily_extract` on the best sources when more detail is needed.

## Tips

- Break complex research questions into multiple focused searches.
- Do not use search when the user provided exact URLs and only wants their content.
- Use `include_raw_content` only when you need full content from result pages; otherwise snippets are cheaper and easier to inspect.
- If results are too broad, add domain filters, date filters, or a more specific query.