Toolso.AI
Toolso.AI
All ToolsCategoriesTrendingLatest ToolsPricingBlog
Toolso.AI
Toolso.AI
Toolso.AI
Toolso.AI

Discover the best AI tools to boost your productivity

GitHubGitHubTwitterX (Twitter)YouTubeYouTubeTikTokEmail

Popular Categories

  • AI Writing
  • AI Image
  • AI Video
  • AI Coding
  • More Categories

Explore

  • Latest Tools
  • Popular Tools
  • More Tools
  • Submit Tool
  • Pricing

About

  • About Us
  • Contact
  • Blog
  • Changelog

Legal

  • Cookie Policy
  • Privacy Policy
  • Terms of Service
  • Refund Policy
© 2026 Toolso.AI All Rights Reserved
Limited timeLimited-time offerFeatured Listing24h priority review · No backlink · 30 days featured$29.90then $59.90Price rises to $59.90 after Oct 31Ends in--:--:--Submit now
  1. Home
  2. All Tools
  3. Research
  4. Exa
Exa interface preview
Exa logo

Exa

Exa gives AI applications and agents search, page-content, answer and deep-research endpoints on top of its own web index, billed per request from prepaid credits with $10 of free credits every month.

ResearchDeveloper ToolsAI Agent#Api#Sdk#Web Search
View Pricing
Saves
Visits
Views
Pricing
Paid
Published
Oct 6, 2026
Domain
exa.ai
Community rating

Used this tool? Rate it

Rate this tool

Exa Product Information

View Pricing
Tool Information
Saves
Visits
Views
Pricing
Paid
Published
Oct 6, 2026
Domain
exa.ai
Community rating

Used this tool? Rate it

Rate this tool

Featured Tools

Related Tools

View Pricing

What is Exa?

Exa is a web search and retrieval API built for AI applications and agents: it combines its own search index with page-content extraction and agentic research endpoints, so software can find sources, pull their text and return grounded, cited output. The Exa search API is called from code, agent frameworks or AI assistants.

The service is operated by Exa Labs Inc., whose privacy policy covers its website, search engine, search API and the Websets tool. The company states that, as of August 2026, its index tracks 1.4 trillion URLs and serves 100 billion pages from across the public web.

The product line has five metered API surfaces: Search (including the slower Deep Search modes), Contents, Answer, Monitors and the asynchronous Exa Agent, plus a separate list-building product called Websets. Exa describes its user base as 500,000 developers, as captured on October 5, 2026. The company says it reached a $2.2B valuation in May 2026. Independent tech coverage reported that Exa Labs had raised $250 million against a $2.2 billion valuation.

Core features

As a web search API, Exa returns ranked results with title, URL and publication date, and the same request can add highlights, full text or a summary of each page. The search type trades latency against research depth.

Search types and latency

TypeDocumented typical latencyIntended use
auto (default)about 1 secondDefault balance of quality and speed
fastabout 450 msLatency-sensitive requests
instantabout 250 msReal-time paths such as autocomplete or voice
deep-liteabout 4 secondsLightweight research and synthesis
deep4–15 secondsMulti-step search and stronger synthesis
deep-reasoning12–40 secondsHarder research tasks

These latency figures are approximate and describe the search itself with cached content; requesting a structured output schema adds a synthesis step, and a low content-age limit adds a fresh page fetch.

Retrieval scope and controls

  • Index coverage: Exa searches the public web and select private data sources, with coverage refreshed continuously.
  • Data categories: source-specific retrieval is available for categories such as companies, people, news and code.
  • Freshness setting: maxAgeHours controls how fresh the extracted content of each result must be; it does not filter results by publication date.

Contents API

Contents works on URLs you already have and returns clean page content, handling JavaScript-rendered pages, PDFs and complex layouts automatically. It can also follow links from each starting URL to crawl subpages.

Answer and Monitors

  • Answer: returns an LLM answer to a question, with citations. The response includes both the generated answer and the sources used to create it, and it can stream tokens as they are generated.
  • Monitors: run Exa searches on a recurring schedule and deliver newly found results to a webhook endpoint.

Exa Agent

Exa Agent is an asynchronous, usage-based endpoint for high-compute work such as list building, enrichment and deep research. A run can return a natural-language answer, schema-validated JSON, field-level grounding, metadata and a cost breakdown.

Websets

Websets builds verified, enriched lists from a natural-language query and a target item count. Exa now points new list-building and enrichment workflows to Exa Agent; the Websets API requires a paid Websets plan, and its credits are separate from Search API credits.

Guide

A first integration follows a short path from key to production call:

  1. Create an API key. Keys are created in the Exa dashboard, and new accounts start with free credits.
  2. Install an SDK or call the REST API. The official SDKs cover Python and JavaScript.
  3. Pick a search type per code path. Use instant or fast where a person is waiting, keep auto as the default, and reserve deep types or Exa Agent for slower research tasks.
  4. Restrict sources only as hard constraints. The includeDomains filter restricts results to trusted sources and accepts full domains, path prefixes and subdomain wildcards. The excludeDomains filter removes domains or paths whose results would be unusable, rather than expressing a preference.
  5. Set freshness only for volatile pages. Most searches should leave the freshness field unset; it is meant for cases where stale page content would be unusable, such as prices, availability or frequently changing pages.

Exa use cases and examples

  • Coding agents: Exa markets search over millions of GitHub repos, docs and Stack Overflow with high-accuracy code retrieval, so generated code can rely on current package and API references. In a customer story on the Exa site, Cognition states that Exa powers all parts of its Devin agent.
  • News and change tracking: Monitors are positioned for following news, competitor announcements, funding rounds, regulatory changes, research publications or any other topic that changes over time.
  • List building and enrichment: Exa Agent can build lists from open-ended criteria and then enrich each result. HubSpot's AI lead says in an Exa-published case study that Exa is necessary for agents on HubSpot to supplement internal data with real-time people and company search.
  • Grounded answers in chat products: Exa returns source URLs and the web content used for retrieval, so an application can show citations and inspect the supporting evidence, while the application and its model remain responsible for how that material is interpreted.

Who is it for

Exa fits teams that call a managed search layer from their own software.

  • Agent and RAG developers who need search, page text and citations inside a tool-calling loop, with latency options from sub-second lookups to slower research modes.
  • Go-to-market teams building account lists: Exa pitches a company data API for finding accounts, enriching a CRM with cited sources and tracking deal signals.
  • Model training and evaluation groups: Exa positions an index of 100B documents and versioned snapshots of the web for pre- and post-training.
  • Inference providers that want to ground the models they serve, which Exa addresses as a web search API for LLMs.
  • Regulated organizations that can sign an Enterprise contract, since Zero Data Retention and HIPAA mode are not part of the free or pay-as-you-go tiers.

It is a weaker fit for people who want a finished consumer search app, for teams that need more than 100 results per call without negotiating Enterprise limits, and for anyone whose traffic originates in the regions Exa blocks.

Platforms

  • REST API and SDKs: the official SDKs search the web, get page contents and get answers with citations.
  • Exa MCP server: the hosted server connects Claude, Cursor, VS Code and other AI tools to Exa's search, code search and research capabilities, and the project is free and open-source.
  • Keyless versus signed-in MCP use: connecting without a key uses Exa's free rate limits; signing in with OAuth or adding an API key moves requests onto your team's plan and limits. Agent runs are usage-based, so the agent tool needs OAuth or an API key.
  • Assistant connectors: the free Starter tier lists MCP server access, a Claude Connector, a ChatGPT plugin and 50+ integrations.
  • Framework integrations: Exa lists integrations with OpenAI, Vercel AI SDK, LangChain, CrewAI, LlamaIndex, Pydantic AI and n8n, among others.
  • Regional availability: to comply with sanctions and trade restrictions, Exa blocks API access from Crimea, Cuba, Iran, North Korea, Russia, Syria, Ukraine and Venezuela, and such requests may receive a Cloudflare block page instead of a normal API error.

Pricing

Exa pricing is pay-as-you-go: there is no subscription and no minimum spend, you load prepaid credits, and each request is charged at the published rate.

Plans

  • Free: $10 in credits on the day you sign up (up to 2,500 Instant searches), reset to $10 on the first of every month, with no payment method required; completing dashboard onboarding adds a one-time $10 bonus for your first team.
  • Free-credit rules: the monthly grant does not accumulate, so unused free credits do not carry into the next month, and only a user's first team receives it.
  • Pay as you go: prepaid credits at a default 10 QPS with 50 active Agent runs. Adding $1,000 in credits within any 30-day window raises the team limit to 25 QPS for 90 days.
  • Enterprise: custom MSA and DPA, Zero Data Retention, SOC 2 Type II and HIPAA, monthly invoicing, SSO and SCIM for the dashboard, custom QPS and up to 1,000 results per search.

Endpoint rates

EndpointBase price (up to 10 results)Each result above 10AI page summaries
Search, instant$4 per 1k requests$1 per 1k results$1 per 1k pages
Search, fast or auto$7 per 1k requests$1 per 1k results$1 per 1k pages
Deep Search, deep-lite or deep$12 per 1k requests$1 per 1k results$1 per 1k pages
Deep Search, deep-reasoning$15 per 1k requests$1 per 1k results$1 per 1k pages
Answer$5 per 1k requests——
Monitors$15 per 1k requests$1 per 1k results$1 per 1k pages
Contents$1 per 1k pages, per content type—$1 per 1k pages

The base price covers the first 10 results of a request, and every result above 10 adds $1 per 1k results. Contents bills each content view separately, so one page returned with both text and highlights counts as two.

Agent pricing

Agent effortPrice
minimal$0.012 per request
low$0.025 per request
medium$0.10 per request
high$0.50 per request
xhigh$1.00 per request
auto (default)metered, default cap $5 per run
ultrametered, default cap $20 per run

Metered runs bill actual usage up to their per-run cap. The usage components are Agent Compute Units at $0.10 per ACU, search tool calls at $0.005 each, email contact enrichment at $0.02 per email and phone enrichment at $0.07 per number. Exa Connect pulls live data from third-party providers through the same integration and is billed on top of Agent usage.

Billing rules

  • Exhausted credits: when a team runs out of credits, requests return 402 Payment Required, and a key that reaches its assigned budget returns the same error.
  • Auto recharge: each automatic top-up can be set from $5 to $10,000, with an optional monthly maximum.
  • Refunds: free credits are never paid out or refunded. The Terms of Service state that subscription payments are nonrefundable with no credits for partially used periods.
  • Grants: startups and education projects can apply for up to $1000 worth of free credits.

Exa alternatives

Tech press coverage has placed Exa in a wave of startups vying to transform search alongside Tavily, TinyFish and Parallel Web Systems. The table compares published entry pricing, but each service meters differently.

ServiceFree allowancePublished search pricingMetering model
Exa$10 in credits every month$4 per 1k instant, $7 per 1k autoPer request, prepaid credits
Tavily1,000 API credits every month$0.008 per credit on pay-as-you-goCredits per request depth
Parallel Search API—$1 per 1K requests (Turbo) to $5 per 1K (Advanced)Per request by mode
Brave Search API$5 in credits every month$5 per 1,000 requestsPer request
  • Tavily meters by credits: a basic search costs 1 API credit and an advanced search costs 2. Its 1,000 monthly free credits come without a credit card.
  • Parallel splits its search API into modes: Turbo targets about 200 ms at $1 per 1K requests for voice, chat and high-volume lookups. Advanced takes about 3 seconds at $5 per 1K requests for background research and deep reasoning agents.
  • Brave Search API lists a capacity of 50 queries per second on its Search plan.

Beyond basic search, Exa's own price list adds Deep Search modes, the asynchronous Agent and separately billed Contents, which the table does not capture.

Limitations

  • Retention controls are Enterprise-only: Zero Data Retention is available on Enterprise plans and enabled per team. It is not currently supported for Answer or Websets.
  • Agent runs under ZDR expire fast: run data is available while the run executes and for up to 10 minutes after it reaches a terminal state, after which the run cannot be retrieved.
  • HIPAA mode is narrow: it supports cached retrieval only, and Search requests must use the instant or fast type and ask for text or highlights rather than summaries.
  • Result caps: Search returns 10 results by default and up to 100 on request, may return fewer, and does not support pagination.
  • Default rate limits: 10 QPS for search, answer and chat completions; 5 QPS for deep search types; 100 QPS for contents; and 5 QPS with 50 active runs for Agent.
  • Latency figures differ by page: the pricing page lists configurable search latency of 180 ms to 1 s. The search documentation instead lists about 250 ms as the typical latency for instant.
  • Crawler boundaries: Exa's crawler respects the Robots Exclusion Protocol, limits request rates per site and does not attempt to bypass logins, paywalls or CAPTCHAs.
  • Output accuracy: the Terms of Service warn that the services may generate output containing incorrect, biased or incomplete information and that output is not a substitute for professional advice.

Privacy and data rights

Exa's privacy policy says Query Data is used to improve its products and technology, including by training and fine-tuning the models that power its services. The same policy states it does not apply to information Exa processes on behalf of business customers as a processor or service provider, which is governed by the customer agreements instead.

  • Personal data in queries: the query fields are not intended for submitting personal information. The policy also says Exa does not actively monitor query data for personal information.
  • License over inputs and outputs: the Terms of Service grant Exa a perpetual, irrevocable license to use User Input and Output as needed to provide, develop and improve its products and services.
  • Security documentation: Exa says it is SOC 2 Type II certified. Its Trust Center hosts the SOC 2 reports, the Data Processing Agreement and other security documentation.

FAQ

Q1. Is Exa free to try?

Yes. New accounts get $10 in free credits that reset monthly, plus a one-time $10 onboarding bonus, with no payment method. The hosted MCP server also works without a key at free rate limits.

Q2. Does Exa use my search queries to train models?

Its privacy policy says Query Data is used for training and fine-tuning its models, while processing for business customers falls under their agreements. Zero Data Retention is Enterprise-only.

Q3. What happens when my credits run out?

Requests return a 402 Payment Required error until credits are added; auto recharge can top up the balance automatically.

Q4. Is Exa available in every country?

No. Exa blocks API access from sanctioned or restricted locations, including Crimea, Cuba, Iran, North Korea, Russia, Syria, Ukraine and Venezuela.

Know a Similar Tool?
If you know other great AI tools, feel free to submit them to us