Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tavily.com/llms.txt

Use this file to discover all available pages before exploring further.

​
Introduction

Claude is Anthropic’s AI assistant designed for reasoning, coding, and research workflows across multiple environments like Claude Desktop, claude.ai, Claude Code, and Claude Cowork. Tavily integrates with the Claude ecosystem in two main ways:
  • As a Connector — a one-click, OAuth-based integration available in Claude Desktop, claude.ai, and Claude Cowork. Built on top of MCP (Model Context Protocol).
  • As a Plugin — a packaged installation for Claude Code that bundles Tavily’s tools and slash commands directly into your terminal workflow.

​
Connector vs. Plugin

ConnectorPlugin
Where it runsClaude Desktop, claude.ai, Claude CoworkClaude Code (terminal), Claude Cowork (desktop)
Install methodOne-click + OAuth in Claude Settings/plugin install tavily (CLI) or Add Plugins in Cowork
AuthOAuth flowTAVILY_API_KEY in ~/.claude/settings.json
Best forChat-based research, everyday Claude usageDeveloper workflows, scripted research, slash-command power users
InvocationAutomatic — Claude picks the tool when neededAutomatic or via slash commands (e.g., /tavily:search)
Pick whichever matches where you use Claude — or use both.

​
Connector

​
Tavily + Claude

Tavily integrates with Claude as an official connector, giving Claude access to:
  • Real-time web search
  • Content extraction from URLs
  • Website crawling and mapping
  • Deep research workflows
Once connected, Claude can automatically use Tavily whenever external information is required.

​
Supported Claude surfaces

Tavily works across the Claude ecosystem:

​
Installation

Onboarding Tavily Connector on Claude
Go to Settings inside Claude.
Click on the Connectors tab.
Search for Tavily and click the + (Connect) button.
Complete the OAuth flow to connect Tavily.
After connecting, go to Configure and enable Allow always (recommended).This allows Claude to automatically use Tavily whenever web search or external data is needed.

​
Tavily tools available

ToolDescription
tavily_searchReal-time web search
tavily_extractExtract clean content from URLs
tavily_crawlCrawl multiple pages from a site
tavily_mapDiscover site structure and URLs
tavily_researchMulti-step deep research workflows
tavily_skillSearch the best skills for your agent

​
How Tavily works inside Claude

Once connected, Tavily runs automatically inside Claude:
  • Claude detects when external data or web search is needed
  • Tavily tools are invoked automatically
  • Results are returned and used in Claude’s response
If Allow always is enabled, everything works seamlessly without the need of manually accepting it.

​
Example use cases

Query: “What are the latest updates in AI this week?” What happens: Claude identifies this as a real-time information request and calls tavily_search.
  • Tavily fetches recent news, blogs, and updates
  • Claude selects the most relevant sources
  • Results are synthesized into a concise summary
Outcome: A current, source-backed overview of the latest AI developments.

​
tavily_extract

Query: “Summarize this article: https://example.com/ai-report” What happens: Claude detects a URL and calls tavily_extract.
  • Tavily extracts clean content from the page
  • Removes boilerplate (ads, navigation, etc.)
  • Returns structured text
Claude then summarizes or analyzes the extracted content. Outcome: A clean, accurate summary of the article without noise.

​
tavily_crawl

Query: “Go through Stripe’s documentation and explain how subscriptions work” What happens: Claude needs multiple pages to answer this.
  • Calls tavily_crawl on the documentation root
  • Tavily traverses linked pages
  • Relevant pages are collected and processed
Claude aggregates information across pages and generates a unified explanation. Outcome: A complete answer built from multiple documentation pages.

​
tavily_research

Query: “Do a deep analysis of the AI chip market and key players” What happens: Claude recognizes this as a complex, multi-step research task.
  • Calls tavily_research (deep research agent)
  • Tavily performs multi-source search, extraction, and synthesis
  • Iteratively refines findings across sources
Claude then compiles a structured, high-quality research report. Outcome: A comprehensive, multi-source analysis rather than a simple summary.

​
Plugin

The Tavily Plugin brings Tavily’s tools directly into Claude’s developer surfaces — Claude Code (terminal) and Cowork (desktop). Install it once and use Tavily via slash commands or let Claude invoke the right skill automatically.

​
Install

​
Prereq: Tavily API key

Add to ~/.claude/settings.json: 
{
  "env": {
    "TAVILY_API_KEY": "tvly-your-key-here"
  }
}
Get a key at tavily.com.

​
Option A — Claude Code (CLI)

/plugin install tavily
Then /clear and Ctrl+C to restart.

​
Option B — Cowork (desktop)

  1. Click the + icon
  2. Add Plugins → Anthropic and Partners
  3. Search Tavily → Add
  4. (Optional) Customize via settings

​
Use (both surfaces)

/tavily:search latest news on EU AI Act
/tavily:research electric vehicle market 2026
/tavily:crawl https://docs.tavily.com
/tavily:extract https://example.com/article
Or just ask Claude naturally — it’ll pick the right skill automatically.

​
Learn more