How large is Exa's index?

Exa reports serving 100 billion documents from its vector DB — described as the world's largest independent index — and tracking 2 trillion URLs, covering the majority of the web.

Is Exa secure and compliant?

Exa holds SOC 2 Type II certification and supports GDPR and CCPA requirements. HIPAA safeguards are available with a BAA, and Enterprise plans offer zero data retention, meaning queries and results are never stored or trained on, plus custom MSA/DPA agreements.

Does Exa integrate with AI agent frameworks?

Yes. Exa offers an MCP server, a Claude Connector, a ChatGPT plugin, and 50+ integrations, all available on the free Starter plan.

Can Exa return structured data?

Yes. Exa can extract enrichments as structured outputs (e.g., company name, CEO, founded year) across 70M+ companies, and the Agent API supports structured outputs with web-grounded citations.

Exa — The web search API built for AI agents

Exa is a web search API built for AI agents, providing realtime search over a large index of public web and private information. It offers multiple endpoints (Search, Contents, Agent, Monitors, Deep Search) with configurable latency from fast (~450ms Lite) to deep research (~10s), token-efficient content extraction, structured outputs with citations, and vertical-specific search across companies, people, news, research, and financial data.

Key Features

Search API with configurable latency

Web search tool calls for agents with configurable latency tiers: Lite (~450ms), Auto (~1s), and Deep (~10s), spanning the full web plus verticals like news, companies, research, people, and financials.

Token-efficient highlights and contents

Contents API returns full page contents or concise excerpts ('highlights') extracted by a specialized model, claiming up to 90% token reduction; up to 25T tokens per week are passed to models as highlights-powered excerpts.

Agent API for deep research runs

Asynchronous deep research runs with structured outputs and grounded citations, with access to leading data providers.

Monitors API for recurring web checks

Identifies fresh events on the web by running searches at a specified cadence (e.g., daily or weekly) and delivering updates via webhooks.

Structured enrichments

Extracts enrichments as structured outputs (e.g., CEO name, founded year) across a database of 70M+ companies.

Multiple search modes per endpoint

Endpoint options include Lite Search ($4/1k requests), Auto Search ($7/1k), Deep Search ($12/1k), Answer ($5/1k), Contents ($1/1k pages per content type), and Monitors ($15/1k).

Agent framework integrations

Includes an MCP server, Claude Connector, ChatGPT plugin, and 50+ integrations available on the free tier.

Enterprise security and compliance

SOC 2 Type II, GDPR, CCPA, and HIPAA support (BAA available), plus zero data retention options where queries and results are never stored or used for training.

How It Works

Sign up and get free credits

Start for free with over $120 in credits per year — $10 credits every month plus a $10 onboarding bonus, with no payment method required.

Call the Search API

Send queries to the Search API and configure latency (Lite, Auto, or Deep), category (full web, news, companies, research, people, financials), and output type (results, answer, or structured output).

Add contents, agents, or monitors

Fetch full page contents or token-efficient highlights, launch asynchronous deep research Agent runs with structured outputs and citations, or schedule recurring Monitors with webhook updates.

Pros & Cons

Pros

Configurable latency from ~450ms (Lite) to ~10s (Deep) in a single API, covering both latency-sensitive agent loops and deep research
Token-efficient highlights can cut tokens passed to models by up to 90%
Strong claimed benchmark results: Exa 3.0 leads on BrowseComp, DSQA, and WideSearch
Generous free tier: $10/month in credits plus a $10 onboarding bonus, no payment method required, with all endpoints and 50+ integrations
Broad index scale: 100 billion documents served, 2 trillion URLs tracked
Strong compliance posture: SOC 2 Type II, GDPR, CCPA, HIPAA (BAA), and zero data retention on enterprise plans

Who It's For

Best for

Developers building AI agents that need realtime web search with configurable latency
Coding agents and assistants (e.g., Cognition uses Exa to power Devin) that need current docs, repos, and changelogs
GTM and enrichment workflows that require structured company/person data at scale
Teams with compliance needs (SOC 2, HIPAA, GDPR, zero data retention) considering enterprise search APIs

Not ideal for

Teams needing a self-hosted or open-source search solution — Exa is a hosted, proprietary API
Very low-volume hobby projects that don't want to monitor per-request costs across multiple endpoints

Use Cases

Coding agents: search millions of GitHub repos, docs, and Stack Overflow with low-latency Lite search inside latency-sensitive agent loops

Chatbots and AI assistants: ground responses in realtime web data with citations

News monitoring: detect fresh events via scheduled searches delivered through webhooks

Enrichment: extract structured data like company CEO and founding year across 70M+ companies

Voice agents: low-latency search calls within conversational loops

People and company search: build lists of matching companies or people (FindAll)

Pricing

Free Starter tier with $10 monthly credits plus a $10 onboarding bonus (no payment method required). Pay-as-you-go Developer tier with per-request billing. Custom Enterprise tier with dedicated support, zero data retention, and custom limits. Individual endpoints: Search from $4/1k requests (Lite), Auto Search $7/1k, Deep Search $12/1k, Answer $5/1k, Contents $1/1k pages per content type, Monitors

Starter

Free

$10 credits every month plus a $10 onboarding bonus, no payment method required. Includes all endpoints, MCP server, Claude Connector, ChatGPT plugin, 50+ integrations, 10 Search QPS, and 50 agent concurrency. Data: web, people, company, an

Developer

Pay as you go

Usage-based billing with no commitment. Everything in Starter plus standard email support, card or bank account payment, SOC 2 Type II, up to 25 Search QPS, and self-serve Exa Connect data providers.

Enterprise

Custom

Everything in Developer plus custom MSA/DPA, zero data retention, SOC 2 Type II and HIPAA, monthly invoicing, SSO/SCIM, organization unified billing, dedicated Slack support, enterprise SLAs, custom QPS and agent concurrency, up to 1,000 re

Search (Lite)

From $4 /1k requests

Web search tool calls for agents with webpage text and highlights; configurable latency from 450ms to 1s.

Search (Auto / Deep)

$7 /1k (Auto); $12 /1k (Deep)

FAQ

Is Exa free to try?

Yes. The Starter plan is free, giving $10 in credits every month plus a $10 onboarding bonus — over $120 in credits per year — with no payment method required. It includes access to all endpoints, the MCP server, Claude Connector, ChatGPT plugin, 50+ integrations, 10 Search QPS, and 50 agent concurrency.

How does Exa pricing work?

After the free tier, pricing is usage-based per request by endpoint: Lite Search from $4/1k requests, Auto Search $7/1k, Deep Search $12/1k, Answer $5/1k, Contents $1/1k pages per content type, Monitors $15/1k, and Agent runs $0.012–$1.00 each. Each additional result above 10 costs $1/1k requests, and AI page summaries cost $1/1k pages. Enterprise plans are custom-priced.

What is Exa used for?

Exa provides web search, contents, deep research, and monitoring APIs for AI applications. Common use cases include coding agents searching docs and GitHub repos, chatbots grounding answers in realtime web data, news monitoring via scheduled webhooks, company/people enrichment with structured outputs, and GTM lead generation.

How fast is the Exa Search API?

Exa