Exa
Exa is the web search API for AI agents that need real-time, structured web data.
Exa is the API I'd pick when search quality on hard queries matters more than shaving the last cent off per-request cost. It leads FRAMES and Tip-of-Tongue, and the token-efficient highlights genuinely change your LLM bill math. The caveat is real: pay-as-you-go adds up fast on high-volume agent loops, so model the per-1k-request math before you commit.
Verified 14d ago · liveness 86/100 · cite: rightaichoice.com/tools/exa
- Coding agents that need fresh docs, repos, and changelogs to ground code generation
- Research assistants running multi-step queries where grounded citations are required
- Sales and GTM teams doing lead enrichment across companies and people at scale
- News and event monitoring pipelines that need webhook delivery on a fixed cadence
- End users who want a graphical search engine to browse results themselves
- Projects that need full raw HTML or rendered page markup rather than excerpts
- High-volume production traffic on a free budget, since Starter caps at $10/month in credits
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Exa if you need a general-purpose web search for humans, require full HTML page retrieval, have a very tight budget, or need extensive custom crawling and indexing controls.
Going past 5 Search QPS on the free Starter tier means you'll need to upgrade to Developer, where you pay per request.
Exa's pay-as-you-go pricing (Search $7/1k requests, Agent from $0.012/run) is competitive for AI agents that need token-efficient, low-latency search. Compared to Perplexity's API (higher per-request cost, less structured output) or Brave Search API (lower cost but less agent-specific features), Exa offers better value for high-volume agentic workloads.
In short
Exa — Exa is the web search API for AI agents that need real-time, structured web data. Best for Coding agents that need fresh docs, repos, and changelogs to ground code generation, Research assistants running multi-step queries where grounded citations are required, Sales and GTM teams doing lead enrichment across companies and people at scale. Free to use.
What's new in Exa
Checked 5 days agoAcross the latest 7 updates: 4 feature updates, 2 launches and 1 changelog entry.
Introducing Exa Snapshot: Versioned View of the Web
Exa Snapshot launched, providing a versioned view of the web for search. Enables querying historical page states.
Dynamic Highlights
Exa published a blog post on Dynamic Highlights, a research preview feature that optimizes context allocation across results.
Dynamic Highlights (research preview)
Dynamic Highlights selects excerpts across the full result set, improving shared context use. Auto showed ~49% better token efficiency and 2.4% higher quality in single-turn RAG.
SOTA Search Over Academic Publications
Exa announced state-of-the-art search over academic publications, covering 350M papers with enriched org and people results.
Publication research expansion
Exa expanded publication research to 350M papers and added enriched organization/people profiles. New category:"publication" replaces research paper; pdf, github, and tweet categories deprecated.
Exa Agent and Exa Connect in MCP
Exa Agent is now available within Exa MCP, usable from Claude, Cursor, and other MCP clients. Exa Connect data sources are available through the Agent flow.
Introducing Exa Connect
Exa Connect launched, giving Exa Agent live access to public and private data via partners including Similarweb, Fiber.ai, Baselayer, Financial Datasets, Affiliate.com, Particle, and Jinko.
What people actually say about Exa — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
68 mentions across 6 sources (Hacker News, YouTube, App Store, Stack Overflow, GitHub, Lemmy) · researched Aug 15, 2026.
Average across the 6 sources that answered — each source counts once, not each post.
- +Sub-180ms instant search latency is a real production win
- +Token-efficient highlights cut context bloat for LLM agents
- +Deep Search gives grounded, cited multi-step research results
- +Structured output extraction for companies and people is handy
- +Free tier with $20 upfront credits lets you test cheaply
- −Community feedback is thin; little real-world reliability data
- −Crawling and deep search costs can climb with volume
- −Token-saving claims need independent verification
- −Tavily and Firecrawl are direct, proven competitors
- −Documentation on latency presets is unclear to some users
- • Crawling (Contents) is billed per page, so high-volume scraping adds up
- • Deep Search costs $12-15 per 1k requests, which is pricier than plain search
- • Monitors (scheduled queries) incur ongoing costs even if results are duplicated
Viability Score
How well maintained and how widely used is Exa? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: September 2026
How we score →Key Features
- Web search API with configurable latency from ~180ms instant to ~1s
- Token-efficient highlights, a trained excerpt model cutting LLM token usage up to 90%
- Agent API for natural-language queries with fixed effort modes (minimal $0.012 to x-high $1.00 per run)
- Deep Search for multi-step agent workflows with grounded citations and structured outputs
- Contents endpoint returning full page text, highlights, and configurable livecrawl policies
- Monitors API detects fresh web events on a cadence and pushes updates via webhooks
- Structured extraction via output_schema across 70M+ companies and 1B+ people
- Category indexes: web, news, companies, people, financials, and scholarly works
- Semantic code, repo, and docs search over millions of GitHub repos
- Exa Connect pulls live data from premium providers (Similarweb, Fiber.ai, Baselayer) in one integration
- MCP server support for Exa Agent (agent_run) and Exa Connect data sources
- Zero data retention; SOC 2 Type II, GDPR, CCPA, HIPAA with BAA available
- Startup and education grants up to $1,000 in free credits
About Exa
Exa is a search API built for AI agents, not humans. Instead of a results page, you get one endpoint that searches the public web and private indexes, pulls full page contents with token-efficient highlights, and returns structured JSON your model can consume directly. It's aimed at developers shipping coding agents, research assistants, enrichment pipelines, and news monitors — the kind of workloads where a stale cache or a page of HTML blows up both accuracy and token budget. The design goal is flexibility across latency and depth. Search with configurable latency from 180ms instant results up to roughly 1s, or push into multi-step agent workflows for grounded answers and citations. Highlights, a specialized excerpt-extraction model, cut token usage dramatically — Exa says up to 25T tokens passed to models each week are concise excerpts rather than raw pages. Output schemas let you extract enrichment fields across 70M+ companies and 1B+ people. The current product line spans a Search API, Contents, Agent, Deep Search, and Monitors, plus category-specific indexes for web, news, companies, people, financials, and scholarly work. Exa Connect (launched June 2026, now available through MCP) bolts premium third-party data providers like Similarweb, Fiber.ai, and Baselayer onto the same agent flow. A July 2026 release pushed state-of-the-art retrieval over academic publications, and Mozilla announced a Firefox and Exa partnership in August 2026. Under the hood: SOC 2 Type II, GDPR, CCPA, HIPAA with a BAA available, zero data retention, and a stated 1.4 trillion URLs tracked. Pricing is pay-as-you-go with a $20 sign-up credit and $10/month ongoing on the free Starter tier. The realistic comparison set is Tavily for agent search and Perplexity or Brave when you just need general web retrieval — Exa competes on accuracy on retrieval benchmarks and on verticals beyond the general web.
Behind the Verdict
Pick Exa when your agent's failure mode is bad context. Coding assistants, research tools, and enrichment pipelines all live or die on whether the retrieved page actually answers the question — that's the gap Exa is built to close, and the benchmark spread over general-purpose search providers is wide enough to matter. Pass when you need an end-user search box with a UI, or when your budget is the binding constraint. Starter gives you $10/month in credits, which is fine for prototyping and nothing like enough for production traffic. At $7 per 1k search requests and $12–15 per 1k Deep Search requests, a busy agent burns real money, and Deep Reasoning and Agent runs cost more still. The closest alternative is Tavily, which is simpler to wire up if you only need basic agent search. Perplexity and Brave are the honest answer if general web retrieval is all you need — Exa's edge shows up on gnarly, multi-hop queries, not on easy lookups. If you're pulling vendor or competitor data, also price Exa Connect against buying those providers directly; the no-separate-contract convenience has a cost on top of Agent usage. Caveats worth knowing before you build. Highlights are excerpts, not full HTML — projects needing raw page markup should look at the Contents endpoint instead. Exa Connect's premium providers are billed on top of Agent usage, and some (ZoomInfo, Crunchbase, Harmonic) are request-only rather than self-serve. There's a startup and education grant worth up to $1,000 in credits; apply before you start paying.
Researching Exa? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Exa actually fits — and what changes day-one when you adopt it.
You want to build a deep research tool that answers multi-part questions with citations.
Outcome: Use Exa Search with type='deep' to get grounded, structured answers with citations in seconds, powered by the Agent API.
You need to enrich leads with company and people data at scale.
Outcome: Use Exa's company and people indexes with output_schema to extract structured data like CEO names and funding, then feed it into your CRM via Stripe or Google Sheets integration.
You need to monitor web events for competitive intelligence.
Outcome: Set up Monitors to watch for fresh content, receive webhook alerts, and automatically update your internal dashboards, with SOC 2 compliance and zero data retention available.
Use Cases
- Power a coding agent with real-time search across docs, repos, and Stack Overflow.
- Build a sales-research tool that turns 'AI startups in healthcare' into a structured company list with CEO names and funding.
- Add neural retrieval to a RAG pipeline for higher quality grounding than keyword search.
- Create monitors that watch for new web content matching specific queries and send webhook alerts.
- Enrich CRM data with company and people intelligence using structured output schemas.
- Build a deep research assistant that synthesizes information from multiple web sources with citations.
- Automate news monitoring and competitive analysis with scheduled searches and summaries.
- Enable a voice agent to fetch up-to-date information in real time.
Models Under the Hood
as of 2026-09-14
Limitations
- Pricing is per-request and per-run (Search $7/1k requests, Agent $0.012–$1.00/run, Contents $1/1k pages), which can stack up on agentic workloads that issue many searches per task.
- Search latency is configurable from 180ms (Instant) up to ~1s, while deep research runs are asynchronous and slower (Auto ~1s, Deep ~10s).
- Free/starter tier provides $20 credits on sign-up plus $10 credits per month, with 10 QPS and 50 agent concurrency; higher QPS (25 self-serve, custom for enterprise) requires paid plans.
as of 2026-08-30
Verification history
We have re-verified Exa 19 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 19 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Exa tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Starter
$0/mo
Ideal for
Developers exploring Exa for prototyping, with $20 free credits on sign-up and $10 monthly credits.
What this tier adds
Free entry point with access to all endpoints, 5 Search QPS, and 3 Agent concurrency, plus MCP server access.
Developer
Pay as you go
Ideal for
Production teams needing pay-as-you-go billing with higher QPS and SOC 2 compliance.
What this tier adds
Adds usage-based billing, 10 Search QPS, 25 Agent concurrency, and self-serve Exa Connect data providers.
Enterprise
Custom
Ideal for
Large organizations requiring custom SLAs, HIPAA compliance, and dedicated support.
What this tier adds
Adds custom MSA/DPA, zero data retention, SSO, monthly invoicing, dedicated Slack channel, and custom QPS limits.
Where the pricing makes sense
The company stage and team size where Exa's pricing actually pencils out — and where peers do it cheaper.
Exa's pay-as-you-go pricing (Search $7/1k requests, Agent from $0.012/run) is competitive for AI agents that need token-efficient, low-latency search. Compared to Perplexity's API (higher per-request cost, less structured output) or Brave Search API (lower cost but less agent-specific features), Exa offers better value for high-volume agentic workloads.
Setup time & first value
How long it actually takes to get something useful out of Exa — broken out by persona, not the marketing-page minute.
Sign up and get an API key in minutes. The dashboard onboarding generates a complete, tested integration snippet for your stack in under one minute. Basic search can be live within 15 minutes; more complex agentic workflows might take a few hours.
Switching to or from Exa
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Bing Search API: Exa offers a migration guide and lower latency for agentic use cases.
- ↗To Perplexity API: If you need a general-purpose search API with a simpler interface, Perplexity is an alternative.
Integrations
Resources & Guides
- Documentationexa.ai
Exa Search API
Exa is an SF-based research lab building perfect search.
- Documentationexa.ai
Search Api Guide For Coding Agents
Full product docs from exa.ai
- Documentationexa.ai
Mcp
Full product docs from exa.ai
- Documentationexa.ai
Llms
Full product docs from exa.ai
- Resourcedashboard.exa.ai
Onboarding
Helpful link from dashboard.exa.ai
Tutorials & Learning
YouTube returned 6 videos for “Exa”, and we withheld 5: 5 could not be judged, because “Exa” is a single word that other videos use for other things. Showing the 1 we can prove is about Exa.
Official links
Featured Head-to-Head Comparisons
Popular in Web Scraping & Search APIs
Spider Cloud
Spider Cloud is a web scraping API that renders, crawls, and searches the web for agents and RAG pipelines.
Mixpeek
Multimodal video search API that finds any timestamped scene by description in your object storage.
Thunderbit
Thunderbit is an AI web scraper Chrome extension that turns any webpage into structured data in one click — no CSS selectors.
Frequently Asked Questions
Categories
Used Exa? Help shape our editorial sentiment research.