Skip to main content
In this guide, you’ll learn how to integrate Tavily—a search API built for large language models (LLMs) and retrieval-augmented generation (RAG)—into your workflows. Tavily provides real-time, up-to-date context by connecting your LLM directly to the web.
The image is a webpage for Tavily AI, promoting a search API optimized for LLMs and RAG, with options to open in GitHub or join the community. It highlights features like in-depth research and intelligent query suggestions.

​
What Is Tavily?

Tavily.com offers a fast, persistent search engine that you can call via API. Think of it as search-augmented retrieval, supplying fresh information your LLM can consume to improve accuracy and relevance.
Tavily’s free tier includes 1,000 API calls per month—perfect for experimentation and small projects.

​
Sign Up and Obtain Your API Key

  1. Go to Tavily.com and create your account.
  2. Navigate to the dashboard to view your plan and API usage.
The image shows a dashboard for Tavily AI, displaying a "Researcher" plan with an API limit of 90 out of 1,000 requests used, along with an API key partially visible.

​
Setting Your API Key

Export your key as an environment variable before running any code.
Operating SystemCommand
macOS/Linuxexport TAVILY_API_KEY="Your Tavily API Key"
Windows PowerShell$env:TAVILY_API_KEY="Your Tavily API Key"
Never commit your TAVILY_API_KEY to public repositories. Treat it like a password!

​
Install the LangChain Community Tool

Make sure you have the required package: 
pip install langchain-community

​
Using TavilySearchResults in Python

Below is a sample script that queries Tavily and prints out the results: 
import os
from langchain_community.tools.tavily_search import TavilySearchResults

# Load your API key from the environment
TAVILY_API_KEY = os.getenv("TAVILY_API_KEY")

# Initialize the Tavily search tool
tool = TavilySearchResults()

# Perform a search query
response = tool.invoke({"query": "When is ICC Men's T20 World Cup 2024 starting?"})

# Inspect the results
print("Number of results:", len(response))
print("First URL:", response[0]["url"])
print("First snippet:", response[0]["content"])

​
Response Structure

Tavily returns a list of dictionaries with the following fields:
FieldDescription
urlThe source link for the search result
contentA short snippet or summary from the linked page
Example output: 
Number of results: 5
First URL: https://www.espncricinfo.com/series/icc-men-s-t20-world-cup-2024-1411166/match-schedule-fixtures-and-results
First snippet: "Get 2024 T20 World Cup schedule, fixtures, scorecard updates, and results on ESPNcricinfo. Track latest match scores, schedule, and results of ICC Men's T20 World Cup 2024."

​
Next Steps

  • Combine multiple snippets into a single context string for your LLM.
  • Explore additional LangChain Community tools for advanced workflows.
  • Experiment with different queries to see how real-time search data improves your applications.

Watch Video