Grounding AI agents
Find relevant pages and return source text or focused highlights that an agent can use as evidence.
Independent tool overview
Exa at a glance
Exa is an AI-focused web search and content-retrieval platform for developers building agents, research tools, RAG systems, monitoring workflows, and structured web-data products.
Visit the official Exa site ↗
- Best for
- AI agents, RAG, research, monitoring, and web-data workflows
- Free access
- $20 signup credit plus $10 in monthly credits
- Starting search price
- $7 per 1,000 Search requests with up to 10 results
- Core interfaces
- REST API, Python and JavaScript SDKs, dashboard, and Websets
- Security
- SOC 2 Type II; custom ZDR and HIPAA options are Enterprise features
- Last reviewed
- August 29, 2026
Overview
What Exa is
Exa gives applications a search engine, crawler, content extractor, cited-answer endpoint, monitoring tools, and an asynchronous research agent through one API platform. Its core Search endpoint can return links alone or bundle clean page text, targeted highlights, and generated summaries with the results.
The product is most useful when a developer needs programmable web retrieval rather than a consumer search interface. Queries can use natural language, domain and date filters, content categories, different speed-versus-depth modes, and freshness controls for cached or live-crawled pages.
Exa can reduce the engineering required to locate and prepare web pages for an AI model, but it does not make web information automatically true or safe to reuse. Teams still need to verify important claims against the cited pages, respect content rights, and distinguish extractive source text from Exa's generated summaries and answers.
Use cases
Who Exa is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Research applications
Run deeper searches, generate cited answers, or use the asynchronous Agent API for multi-step research.
Content extraction
Turn known URLs, JavaScript-rendered pages, PDFs, and linked subpages into cleaner model-ready content.
Web monitoring
Track recurring topics or entities and send newly discovered results into downstream workflows.
List building and enrichment
Use Websets or Agent workflows to discover entities, verify criteria, and enrich records from the public web.
Capabilities
Core Exa features
Multiple search modes
Choose Auto, Instant, Fast, Deep Lite, Deep, or Deep Reasoning modes to balance latency, breadth, and reasoning effort.
Search controls
Filter by included or excluded domains, crawl date, publication date, and categories such as news, research papers, companies, people, or financial reports.
Text, highlights, and summaries
Return clean page text, query-focused excerpts, or LLM-generated summaries, including schema-constrained summary output.
Contents API
Extract known URLs and optionally crawl targeted subpages, with freshness settings that control cached versus live retrieval.
Answer API
Produce a direct search-grounded response alongside cited search results for conversational and question-answering products.
Agent API
Run asynchronous multi-step research and enrichment tasks with cited natural-language or structured output. This API is currently marked beta.
Monitors
Schedule repeated searches to discover new results over time and connect them to automated workflows.
Websets
Build and enrich lists of companies, people, or other entities from a natural-language description and verification criteria.
Process
How the Exa workflow works
Step 1
Choose the smallest suitable endpoint
Use Search when starting from a query, Contents for URLs you already know, Answer for a direct cited response, and Agent for longer multi-step work.
Step 2
Constrain the search
Add trusted domains, exclusions, publication windows, categories, and an appropriate search mode instead of sending an unconstrained prompt.
Step 3
Request only the content you need
Use highlights for focused evidence, full text for deep analysis, and generated summaries only when their added convenience justifies additional cost and validation.
Step 4
Validate the evidence
Open the cited pages, check dates and authorship, compare sensitive claims across sources, and reject results that do not support the conclusion.
Step 5
Measure cost and quality
Track searches, extra results, extracted content types, live crawling, and agent effort separately, then test retrieval quality on representative queries.
Cost
Exa pricing and free plan
Exa uses usage-based pricing. Search, content extraction, deeper search modes, answers, monitors, and Agent runs can each add cost, so production budgets should be modeled from the exact request shape.
Free credits
$0 to start
Developer access with introductory and recurring usage credits.
- $20 in signup credits
- $10 in credits each month
- Access to web search, page text or highlights, and configurable search latency
Search and Contents
From $7 per 1,000 Search requests
Usage-based retrieval for search results and page content.
- Search costs $7 per 1,000 requests with up to 10 results
- Each result above 10 adds $1 per 1,000 requests per extra result
- Text, highlights, and summaries are each billed at $1 per 1,000 pages when chargeable
- Content requested with Search is included for up to 10 results
Deep, Answer, and Monitors
$5–$15 per 1,000 requests
Higher-level retrieval and recurring discovery endpoints.
- Deep Search costs $12 per 1,000 requests
- Deep Reasoning Search costs $15 per 1,000 requests
- Answer costs $5 per 1,000 requests
- Monitors cost $15 per 1,000 requests
Agent
From $0.012 per run plus usage
Beta asynchronous research and enrichment priced by effort and actions.
- Minimal effort is listed at $0.012 per run, Low at $0.025, Medium at $0.10, High at $0.50, and X-high at $1
- Agent search calls are listed at $0.005 each
- Email enrichment is listed at $0.02 per email and phone enrichment at $0.07 per number
- Agent pricing is beta and should be confirmed in the dashboard before budgeting
Enterprise
Custom
Custom scale, support, security, and contracting.
- Volume discounts and up to 1,000 results per search
- Custom throughput, indexes, onboarding, support, SLAs, and MSAs
- Zero Data Retention and customized security options
- Startup and education grants may provide $1,000 in credits
Pricing checked . Check current pricing at the source ↗
Assessment
Exa strengths and limitations
Where it stands out
- Combines search, crawling, content preparation, answers, monitoring, and research workflows in one developer platform
- Offers precise controls for domains, dates, categories, result count, latency, and content freshness
- Can return compact highlights when full-page text would waste model context
- Provides source URLs and citations that make downstream verification possible
- Supports both low-latency retrieval and deeper asynchronous research
- Enterprise buyers can request ZDR and other customized security arrangements
What to consider
- Search ranking, generated summaries, and generated answers can all be incomplete or wrong; citations must still be checked
- Summary output is LLM-generated, while text output is extractive, so the two should not be treated as equally reliable evidence
- Cached content, crawl dates, and reported publication dates may not match the current state of a page
- Costs can compound across search requests, extra results, content modes, live retrieval, and Agent actions
- The Agent API is beta, its pricing may change, and default concurrency is limited
- Published default rate limits include 10 requests per second for Search and 100 for Contents; higher-volume systems may need Enterprise capacity
- Content moderation is not enabled by default on Search requests
- Developers remain responsible for robots directives, site terms, copyright, paywalls, and permission to store or republish retrieved content
- Person and contact enrichment can create privacy, consent, and legal obligations depending on the data and jurisdiction
- ZDR, HIPAA-oriented configurations, and customized security controls require an Enterprise discussion rather than being standard self-serve guarantees
Compare
Exa alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Data Analysis
Brave Search API
Consider Brave when you want a more conventional independent web-search API and its associated search datasets.
Explore Brave Search API →Data Analysis
Perplexity
Consider Perplexity when the primary need is an end-user research and answer experience rather than a retrieval infrastructure layer.
Explore Perplexity →Questions
Exa FAQs
What is Exa AI used for?
Exa is used to add web search, page extraction, cited answers, monitoring, research, and entity enrichment to AI agents and other software.
Is Exa free?
Exa provides $20 in signup credits and $10 in recurring monthly credits. Ongoing production usage is billed by endpoint and request shape.
How much does Exa Search cost?
The current public price is $7 per 1,000 Search requests with up to 10 results. Additional results and chargeable content extraction can add cost.
What is the difference between Search and Contents?
Search begins with a query and finds relevant pages. Contents begins with URLs you already know and extracts their text, highlights, summaries, or linked subpages.
Are Exa summaries copied from the source?
No. Exa documents its text output as extractive source content and its summary output as LLM-generated. Generated summaries should be verified against the cited page.
Does Exa always crawl a page live?
No. Contents can use cached data or live crawling depending on availability and the freshness setting. A zero-hour freshness setting requests live retrieval, while a cache-only setting avoids it.
Does Exa support zero data retention?
Exa offers Zero Data Retention as an Enterprise security arrangement across its search products. Self-serve users should not assume their account has ZDR without a matching agreement.
Is Exa's Agent API production-stable?
The Agent API is currently documented as beta. Teams should expect possible interface, limit, and pricing changes and test failure handling before depending on it.
Bottom line
Our Exa verdict
Exa is a strong option for developers who want one API family for AI-native web search, extraction, cited answers, monitoring, and deeper research. Its best value comes from carefully choosing the endpoint and content mode for each task; teams that treat every query as deep research can quickly add latency and cost. It is retrieval infrastructure, not a substitute for source evaluation, licensing checks, or human review of important conclusions.
Visit Exa website ↗Similar Data Analysis Tools
View all Data Analysis AI tools →
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.