🎉 Tavily raises $25M to power the Internet of AgentsRead more

Features Use Cases Pricing Enterprise Careers Blog Docs

Operational

Connect Your
Agent
to the Web

Powering the Internet of Agents with fast, secure
and reliable web access APIs.

Talk to an Expert

Go to Playground

Trusted by 700k+ developers around the world

Features

Introducing
our Search API

Unleash production-ready agents with a search API designed for LLMs.
Delivering fast and relevant results that reduce hallucinations and make agents more powerful.

from tavily import TavilyClient
tavily_client = TavilyClient(api_key="tvly-YOUR_API_KEY")
response = tavily_client.search("Who is Leo Messi?")
print(response)

const { tavily } = require('@tavily/core');
const tvly = tavily({ apiKey: "tvly-YOUR_API_KEY" });
tvly.search("Who is Leo Messi?")
  .then(results => console.log(results));

curl -X POST 'https://api.tavily.com/search' \  
  -H 'Content-Type: application/json' \ 
  -H 'Authorization: Bearer tvly-YOUR_API_KEY' \ 
  -d '{"query": "Who is Leo Messi?"}'

Plug and play

Simple setup with just two clicks to sign up, get your API key,
and integrate seamlessly into your application.

Read Docs

Real-time web access

Fast, reliable access with high rate limits, accurate results,
and relevant content snippets optimized for AI processing.

API Playground

Approach

Enhance Your
Agentic Workflow

Supercharge your reports with real-time, comprehensive data from across the web, ensuring accuracy and relevance.

Blazing fast and production-ready.

Tavily is the web layer that supplies enterprise-grade agents with
fast, real-time web data through simple, agent-first APIs.

Learn More Contact Us

Benchmarks

Web Search Driven by Research

About this benchmark

OpenAI’s SimpleQA benchmark measures the ability of language models to answer short, fact-seeking questions. It uses a curated set of factual queries designed to test how accurately a model can retrieve and synthesize factual information. Read more about the SimpleQA benchmark here.

Methodology

We ran the full SimpleQA dataset on each provider and used GPT-4.1 to generate predicted answers from the documents retrieved by that provider. These predictions were then graded with SimpleQA’s classifier.
The final accuracy was calculated by dividing the number of correct answers by the total number of questions.

To maintain fairness and ensure documents of similar length were compared across providers, we set include_raw_content=False (default), contents.highlights=True, and max_tokens_per_page=512 for Tavily, Exa, and Perplexity Search, respectively. In addition, we retrieved a maximum of 10 documents per query across all providers.

Resources

Check out our collaboration
with DeepLearning.AI

Testimonials

Tavily is trusted by AI
leaders around the world

We've been using Tavily's Search API to power Athena, our Enterprise AI Data Analyst. The ability to retrieve accurate, real-time information tailored for AI agents has significantly enhanced Athena’s research capabilities. Tavily's focus on delivering factual and explicit results aligns perfectly with our commitment to providing trustworthy and sourced insights to our enterprise clients.

Brendon Geils

Co-Founder of Athena Intelligence

Accurate and up-to-date knowledge has become a key component of building many types of LLM-powered apps. Tavily's reliable search APIs have made them an essential partner for LangChain, and our open-source integration with them has been a huge boon to the community as a whole. It's always a pleasure working with the Tavily team!