skills.rest
agnivon
agnivonCOMMUNITY
Explore all skills by this author →
Website𝕏@agnivonGitHub

tavily-best-practices

Implement Tavily API integrations for web search, extraction, crawling, and research in agentic workflows.

Updated Jun 7, 2026
Search Tags:#research#rag#web search#ai agents#web scraping#api integration#tavily
One-click install
npx skills add https://github.com/agnivon/viral_thread_generator --skill tavily-best-practices-agnivon
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: tavily-best-practices
Source: https://github.com/agnivon/viral_thread_generator/tree/main/.agents/skills/tavily-best-practices
Command: npx skills add https://github.com/agnivon/viral_thread_generator --skill tavily-best-practices-agnivon

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Developers building AI agents, RAG systems, or research pipelines often struggle to integrate web search and content extraction correctly, leading to wasted API credits, context window overflow, and unfocused results. This Skill provides production-ready guidance for the Tavily API so coding assistants implement integrations correctly the first time. ## Core Features & Use Cases - Method Selection Guidance: Clear decision tables for choosing between search(), extract(), crawl(), map(), and research() based on your task. - In-Depth API References: Detailed documentation covering query optimization, search depth tradeoffs, domain filtering, chunk-based extraction, crawl scoping, and structured research output. - Framework Integrations: Ready-to-use patterns for LangChain, LlamaIndex, CrewAI, Pydantic AI, Agno, Vercel AI SDK, Anthropic tool calling, OpenAI function calling, and Google ADK. - Use Case: You are building an agent that researches competitors. The Skill guides you to use research() with the pro model for multi-domain analysis, or to chain search() with score-based filtering and extract() with query reranking for a custom pipeline. ## Quick Start Ask your coding assistant to implement a Tavily web search integration for your agent following the tavily-best-practices skill guidelines.

Frequently Asked Questions about tavily-best-practices

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I add web search to an AI agent with Tavily?▼

Install tavily-python or @tavily/core, initialize the client with your TAVILY_API_KEY, and call client.search() with a query under 400 characters. Use search_depth="advanced" for precision queries and async clients for parallel searches in agentic systems.

What is the difference between Tavily search, extract, crawl, and research?▼

search() returns ranked web results for a query, extract() pulls content from up to 20 specific URLs, crawl() traverses an entire site for full content, and map() discovers URLs only. research() is an end-to-end AI research agent that synthesizes a cited report.

How do I prevent context window overflow when extracting web content?▼

Pass a query parameter with chunks_per_source (1-5) to extract() or instructions with chunks_per_source to crawl(). This returns only the most relevant chunks of up to 500 characters each instead of full page content.

Does Tavily work with LangChain and CrewAI?▼

Yes. Use the official langchain-tavily package for Search, Extract, Map, Crawl, and Research tools; the older langchain_community Tavily tool is deprecated. CrewAI provides TavilySearchTool and TavilyExtractTool via the crewai-tools package.

When should I use the Tavily research API instead of building my own pipeline?▼

Use research() when you want end-to-end research with automatic source gathering and cited synthesis without orchestrating searches yourself. Choose model="mini" for narrow questions, "pro" for multi-domain analysis, and build a custom search-plus-extract pipeline when you need full control.

Why do Tavily crawls take so long or return irrelevant pages?▼

Crawl time grows exponentially with max_depth, so start with max_depth=1 and max_breadth=20. Add instructions for semantic focus, use select_paths and exclude_paths regex filters, and always set a limit to prevent runaway crawls and unexpected credit usage.

Related Skills

Explore More Software Engineering Skills→
fcakyonfcakyon
1.0k

tavily-usage

Search the web and extract content from URLs via Tavily APIs.

Community
Intermediate
kreuzberg-devkreuzberg-dev
9.2k

picking-a-format

Maps document consumers to the correct Xberg CLI format and content-format flags.

Official
Basic
hainamchunghainamchung
54

rag-architect

Design and implement Retrieval-Augmented Generation architectures with hybrid search and vector databases.

Community
Advanced
Zedit42Zedit42

web-scraping-pro

Extracts content from JavaScript-heavy and protected websites using multiple fallback methods.

Community
Advanced