@font-face {
  font-family: 'Reckless';
  src: url('https://exa.ai/search/fonts/RecklessTRIAL-Medium.woff2');
  font-weight: normal;
  font-style: normal;
  font-display: swap;
}

@font-face {
    font-family: 'ABCDiatype';
    src: url("https://exa.ai/search/fonts/ABCDiatype-Regular.otf") format("opentype");
}

@font-face {
  font-family: 'SuisseIntlMono';
  src: url("https://exa.ai/search/fonts/SuisseIntlMono-Regular-WebTrial.woff2") format('woff2');
}

h1, h2, h3 {
  font-family: 'Reckless';
  letter-spacing: normal !important;
  margin-bottom: 1rem !important;
}

p {
    font-family: 'ABCDiatype';
}

.card {
  border-radius: 15px;
  overflow: hidden;
  margin-bottom: 1rem !important;
}

pre {
  margin-top: 0.5rem !important;
  margin-bottom: 0.5rem !important;
}

section,
.section,
.content section {
  margin-bottom: 1rem !important;
}

h1, h2, h3, h4 {
  margin-top: 0.5rem !important;
  margin-bottom: 0.5rem !important;
}

hr {
  margin-bottom: 1.5rem !important;
  margin-top: 1.5rem !important;;
}

.anchor-icon-wrapper {
  border-radius: 50% !important;
}

.anchors-icon {
  border-radius: 50% !important;
}

.py-6 {
  padding-top: 0.5rem;
  padding-bottom: 0.5rem;
}

.mt-4 {
  margin-top: 0.5rem;
}

.text-primary-light {
  color: #FFFFFF !important;
}

.border-primary-light {
  border-color: #FFFFFF !important;
}

.border-primary {
  border-color: #FFFFFF !important;
}

.card-title {
  font-family: 'ABCDiatype' !important;
  font-size: 1.2rem !important;
  padding-top: 0.5rem !important;
}

.mt-2 {
  margin-top: .5rem;
  margin-bottom: .5rem;
}

/* Reduces spacing between title and text in steps */
.w-full.overflow-hidden.pl-12.pr-px p {
  margin-top: 0 !important;
  margin-bottom: 0rem !important;
  padding-top: 0.5rem !important;
}

/* Reduces spacing after code blocks */
.not-prose.rounded-xl {
  margin-bottom: 1rem !important;
} 

hr {
  margin-bottom: 1rem !important;
  margin-top: 1rem !important;
}

.mt-8 {
  margin-top: 1rem !important;
}

.mt-10 {
  margin-top: 1rem !important;
}

.max-h-\[300px\] {
  max-height: 500px !important;  
}

.api-button {
  align-items: center;
  background: rgb(var(--primary));
  color: var(--white, #fff);
  display: flex;
  font-family: var(SuisseIntlMono);
  text-transform: uppercase;
  font-size: 16px;
  font-style: normal;
  font-weight: 400;
  gap: 10px;
  height: 40px;
  justify-content: center;
  line-height: 27px;
  padding: 14px 32px;
  text-decoration: none;
  transition: 0.1s ease-in-out;
  border: 2px solid var(--primary);
  word-spacing: 8px;
  letter-spacing: 1px;
}

.api-button:hover {
  background-color: #000000;
}


Model Capabilities & Recommendations

Latency Benchmarks

Learn about the models available in the Research API

Models

Overview

Welcome to Exa

Make your first request to one of Exa's API endpoints

Quickstart

The search endpoint lets you intelligently search the web and extract contents from the results. 

By default, it automatically chooses between traditional keyword search and Exa's embeddings-based model, to find the most relevant results for your query.

Search

Get the full page contents, summaries, and metadata for a list of URLs.

Returns instant results from our cache, with automatic live crawling as fallback for uncached pages.

Get contents

Find similar links to the link provided and optionally return the contents of the pages.

Find similar links

Get an LLM answer to a question informed by Exa search results. Fully compatible with OpenAI's chat completions endpoint - docs [here](https://docs.exa.ai/reference/chat-completions).
`/answer` performs an Exa search and uses an LLM to generate either:
1. A direct answer for specific queries. (i.e. "What is the capital of France?" would return "Paris")
2. A detailed summary with citations for open-ended queries (i.e. "What is the state of ai in healthcare?" would return a summary with citations to relevant sources)

The response includes both the generated answer and the sources used to create it. The endpoint also supports streaming (as `stream=True`), which will returns tokens as they are generated.


Answer

OpenAPI Specification

Learn how to build your first Retrieval Augmented Generation (RAG) system with Exa and OpenAI. 

RAG with Exa and OpenAI

How to use Exa's integration with LangChain to perform RAG.

RAG with LangChain

Enhance your OpenAI chat completetions with a simple Exa wrapper that handles search, chunking and prompting.

OpenAI Exa Wrapper

Learn how to add Exa retrieval capabilities to your CrewAI agents.

CrewAI agents with Exa

A quick-start guide on how to add Exa retrieval to a LlamaIndex Agent Application.

RAG with LlamaIndex

Learn to use OpenAI's tool call feature with Exa's Search Integration

Tool calling with GPT

Using Claude's "Tool Use" Feature with Exa Search Integration.

Tool calling with Claude

Use Exa's /chat/completions endpoint as a drop-in replacement for your OpenAI chat completions code. This will send queries to Exa's `/answer` endpoint

OpenAI Chat Completions

Exa's web search can now be used with OpenAI's new Response API. This lets you add live web search to your AI app using OpenAI's function calling feature.

OpenAI Responses API

Exa is a novel search engine that utilizes the latest advancements in AI language processing to return the best possible results.

How Exa Search Works

We spend a lot of time and energy creating a high quality, curated index.

The Exa Index

Contents retrieval with Exa API

This page explains some of the available feature functionalities of Exa and some unique ways you might use Exa for your use-case

Exa's Capabilities Explained

FAQs

Crawling Subpages with Exa

Exa LiveCrawl

Details on Team structure and account management for the Exa platform

Setting Up and Managing Your Team

Rate Limits

Enterprise Documentation & Security 

Exa MCP

A live demo that detects hallucinations in content using Exa's search.

Hallucination Detector

Writing Assistant

Chat app

Company researcher

Make your very own Hacker News powered by Exa

Hacker News Clone

Learn how to build an AI-powered news summarizer that searches and summarizes recent articles using Exa and GPT.

Building a News Summarizer

Learn how to build an AI-powered system that identifies and verifies claims using Exa and LangGraph.

Building a Hallucination Checker

Using Exa to enable retrieval-augmented generation.

RAG Q&A

Example project using the Exa Python SDK.

Company Analyst

Exa Researcher - JavaScript

Exa Researcher - Python

Recruiting Agent

Phrase Filters: Niche Company Finder

Tutorial for simple Exa searches on our front-end.

Job Search with Exa

Build a Retrieval Agent with LangGraph

Using Exa with instructor to generate structured outputs from web content.

Structured Outputs with Instructor

LangChain Docs

LlamaIndex Docs

CrewAI Docs

IBM WatsonX

Enumeration of methods and types in the Exa Python SDK (exa_py).

Python SDK Specification

TypeScript SDK Specification

Some common code you might want to use - don't miss the TypeScript tab below!

Python and TS Cheat Sheets

Our goal is to help you find anything you want on the web, no matter how complex.

Welcome to Websets

Welcome to the Websets Dashboard! Find anything you want on the web, no matter how complex.

Get started

Here are some examples for things to search for, to get you started!

Example queries

The Websets API helps you find, verify, and process web data at scale to build your unique collection of web content.

How It Works

Automate in-depth web research and receive structured JSON results with citations.

Welcome to Research

A deep-dive into the agents and lifecycle behind every Research task.

Copy-paste request bodies to jump-start your integration.

Examples

Frequently asked questions about the Research API

The Research API automates end-to-end web research — searching, reasoning, and structuring the results so you don't have to.

The Research API is chat compatible and can be used with the OpenAI chat completions interface.

Chat Completions

The /contents endpoint now returns detailed status information for each URL instead of HTTP error codes, providing better visibility into individual content fetch results.

Contents Endpoint Status Changes

Auto search, which intelligently combines Exa's proprietary neural search with traditional keyword search, is now the default search type for all queries.

Auto search as Default

Discord

Blog

Documentation

Integrations

SDKs

Websets

Research

Changelog

API Dashboard

Exa Search

Create a long-running, multi-step research task that explores the web, gathers sources, synthesizes findings, and returns structured JSON results with citations.

Submit your request to `POST /research/v0/tasks` with natural-language `instructions` and an `output.schema` describing the JSON you want back. If you do not provide an output schema, by default a detailed markdown report will be produced. The API responds immediately with a unique task `id` that you can use to poll the task or list tasks.

**Output Schema Requirements:**  
- Must be a valid JSON Schema  
- Root type must be an `object`  
- All root-level fields must be in the `required` array  
- Must include `"additionalProperties": false` at the root level


Create a task

Retrieve the status and results of a previously created research task.

Use the unique task `id` returned from `POST /research/v0/tasks` to poll until the task is **completed** or **failed**. When the task is complete, the response includes your structured JSON output and citations.


Get a task

Retrieve a paginated list of your research tasks along with their current status and metadata. This endpoint allows you to monitor many tasks at once instead of polling each task individually.

The response follows a same cursor-based pagination pattern. Pass the `limit` parameter to control page size (max 200) and use the `cursor` token returned in the response to fetch subsequent pages.


List tasks

Here's how to create enrichments (also known as Adding Columns). 

Creating Enrichments 

Explore your Websets matched results, view summaries, criteria justification 

Exploring your results 

Adding and Managing Your Team Members in Websets

Here's how to prompt your query in Websets

Prompting Websets

Here's how to share or download your results and enrichments.

Downloading and Sharing Your Results

Creates a new `Stream` to continuously keep your Websets updated with fresh data.

Streams automatically run on your defined schedule to ensure your Websets stay current without manual intervention:

- **Find new content**: Execute `search` operations to discover fresh items matching your criteria
- **Update existing content**: Run `refresh` operations to update items contents and enrichments
- **Automated scheduling**: Configure `frequency`, `timezone`, and execution times

Create a Stream

Get Stream

Update Stream

Delete Stream

List Streams

Create a Webhook

Get a Webhook

Update a Webhook

Delete a Webhook

List webhooks

List all attempts made by a Webhook ordered in descending order.

List webhook attempts

Learn how to securely verify webhook signatures to ensure requests are from Exa

Verifying Signatures

Learn about the events that occur within the Webset API

Types

Event Types

List all events that have occurred in the system.

You can paginate through the results using the `cursor` parameter.

List all Events

Get a single Event by id.

You can subscribe to Events by creating a Webhook.

Model	p50 Time (seconds)	p90 Time (seconds)
exa-research	20	40
exa-ultra	35	80

Getting Started

API

​Model Capabilities & Recommendations

​Latency Benchmarks

Model Capabilities & Recommendations

Latency Benchmarks