> ## 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.

# Claude Connectors

> Use Tavily as an official Claude Connector to enable real-time web search, extraction, crawling, and research across Claude apps.

## Introduction

[Claude](https://claude.ai/) 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.

With **Connectors**, Claude can securely integrate with external tools using [MCP](https://modelcontextprotocol.io/docs/getting-started/intro) (Model Context Protocol), enabling real-time data access and extended capabilities.

***

## 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:

* [Claude Cowork](https://www.anthropic.com/product/claude-cowork)
* Claude Code - Through Claude Desktop, Alternatively if you want to use Tavily through Claude Code terminal, follow [this](https://docs.tavily.com/documentation/mcp#connect-to-claude-code).
* [claude.ai](https://claude.ai/)
* [Claude Desktop](https://support.claude.com/en/articles/10065433-installing-claude-desktop)

***

## Installation

<Frame>
  <img src="https://mintcdn.com/tavilyai/E8iWH_0lGYwtARGq/images/connectors_claude_tavily.gif?s=db789895957834a778b9b8aed2254307" alt="Onboarding Tavily Connector on Claude" width="1092" height="720" data-path="images/connectors_claude_tavily.gif" />
</Frame>

<AccordionGroup>
  <Accordion title="Step 1: Open Claude settings">
    Go to **Settings** inside Claude.
  </Accordion>

  <Accordion title="Step 2: Navigate to Connectors">
    Click on the **Connectors** tab.
  </Accordion>

  <Accordion title="Step 3: Add Tavily">
    Search for **Tavily** and click the **+ (Connect)** button.
  </Accordion>

  <Accordion title="Step 4: Authenticate via OAuth">
    Complete the OAuth flow to connect Tavily.
  </Accordion>

  <Accordion title="Step 5: Configure permissions">
    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.
  </Accordion>
</AccordionGroup>

***

## Tavily tools available

| Tool             | Description                           |
| ---------------- | ------------------------------------- |
| tavily\_search   | Real-time web search                  |
| tavily\_extract  | Extract clean content from URLs       |
| tavily\_crawl    | Crawl multiple pages from a site      |
| tavily\_map      | Discover site structure and URLs      |
| tavily\_research | Multi-step deep research workflows    |
| tavily\_skill    | Search 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

### tavily\_search

**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”](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.

***

## Learn more

* Claude Connectors Directory - [https://claude.com/connectors](https://claude.com/connectors)
* Use Tavily with Anthropic SDK - [https://docs.tavily.com/documentation/integrations/anthropic](https://docs.tavily.com/documentation/integrations/anthropic)