Exa
Contact SalesWeb search API built for AI agents
By Tanmay Verma, Founder · Last verified 20 Jun 2026
5.7k views
Added 4/27/2026
84/100Safe BetIn short
Exa — Web search API built for AI agents. Best for AI agents needing real-time web data for coding, research, or automation, Developers building LLM-powered applications that require structured search results, Lead generation and company enrichment for sales intelligence tools. Free to start; paid plans from $0.0252/mo.
Affiliate disclosure: We earn a commission when you use our links. Editorial picks are independent. How we choose.
Is Exa actually worth it?
LiveSee what real users actually say. We scan live discussions, reviews and complaints across the web and hand you an honest verdict — in under a minute.
3 free scans · no card needed · downloadable report
Editorial Verdict
Best for
AI agents needing real-time web data for coding, research, or automationDevelopers building LLM-powered applications that require structured search resultsLead generation and company enrichment for sales intelligence toolsMarket research applications needing up-to-date competitive dataTeams optimizing LLM token usage with pre-summarized web excerpts
Not ideal for
General-purpose web search for end-users with a graphical interfaceProjects requiring full HTML page retrieval without summarizationBudget-constrained teams needing a free or low-cost search API (pricing is contact-based)Applications that need extensive custom crawling and indexing controlsSimple query-and-response use cases where structured output is unnecessary
Exa is a top-tier search API for AI agents, offering unmatched accuracy and token efficiency. If you need fast, structured web data for your agent, Exa is the clear choice over generic search APIs.
Last verified: June 2026
Behind the Verdict
Exa has evolved from a niche search API into a full-fledged agentic search platform. The March 2026 pricing update makes it more accessible, with search with contents now at $7/1k requests and summaries at $1/1k. The new Exa Agent API (June 2026) allows natural-language queries with effort modes, competing with specialized research agents. However, Exa remains a paid, contact-based service, so budget-constrained teams may find cheaper alternatives like Brave Search API. For AI agents needing reliable, structured web data, Exa's highlights feature can reduce token costs by up to 90%, a game-changer for LLM applications. The deprecation of legacy endpoints (April 2026) means existing integrations may need updates. Exa's benchmark leadership (FRAMES, Tip-of-Tongue, Seal0) reinforces its accuracy advantage over Perplexity and Brave. If your use case is general web search for end-users, stick with Google or Bing. But for agent workflows, Exa is the best-in-class option.
Skip Exa if Skip Exa if you need a free or low-cost search API for casual browsing or scraping — Brave Search or Perplexity's free tier are cheaper alternatives.
Latest from Exa
Updated 2 days agoAcross the latest 7 updates: 3 feature updates, 2 launches, 1 pricing change and 1 changelog entry.
LaunchChangelog·20 days agoNewest
Exa Agent Released
Exa Agent API launched: frontier web research agents via API, supporting natural-language queries, effort mode, and outputSchema.
ChangelogChangelog·Apr 15
API Deprecations: /research endpoint, legacy fields removed
Legacy /research endpoint and response fields resolvedSearchType, highlightScores deprecated. startCrawlDate / endCrawlDate ignored.
PricingChangelog·Mar 1
Exa Pricing Update: lower prices, free contents for first 10 results
Search with contents $7/1k requests; summaries $1/1k; Exa Deep $12/1k; Deep Reasoning $15/1k. Contents endpoint $1/1k pages.
FeatureChangelog·Mar 1
Exa Deep Revamp: faster, cheaper, structured outputs
Deep search now 4-12s; deep-reasoning type added. 20% price cut. outputSchema with field-level grounding.
LaunchChangelog·Mar 1
Exa Monitors Launched
Scheduled searches with webhook delivery, deduplication, structured output, and flexible intervals (min 1 hour).
FeatureChangelog·Feb 1
content freshness, maxCharacters for highlights, MCP updates
maxCharacters replaces numSentences for highlight control. Content freshness and MCP server improvements.
FeatureChangelog·Feb 1
Exa Instant Search: sub-150ms neural search
New search type 'instant' with sub-150ms latency, designed for real-time apps like chat and voice AI.
Viability Score
84/100
Safe BetHow likely is Exa to still be operational in 12 months? Based on 4 signals — momentum (how recently it shipped), wrapper dependency, revenue model, and web presence.
momentum
100
funding runway
70
website health
90
github activity
45
wrapper dependency
100
Last calculated: June 2026
How we score →About Exa
Exa is a web search API designed specifically for AI agents, offering fast and accurate search, crawling, and research capabilities. It provides one API for search, crawling, and research agents, making it ideal for developers building AI-powered applications that require real-time web data. Key features include token-efficient highlights that reduce token usage by up to 90%, structured outputs from web content (e.g., company data enrichment across 70M+ companies), and low-latency responses as fast as 150ms. Exa also offers a Deep Agent for complex multi-step research, Monitors for continuous data updates, and the newly launched Exa Agent for natural-language queries with effort modes. Trusted by teams like HubSpot and Cognition (Devin), Exa leads benchmarks on FRAMES, Tip-of-Tongue, and Seal0. Compared to traditional search APIs, Exa is purpose-built for AI agent workflows with features like highlights and structured data extraction. Recent pricing updates (March 2026) introduced lower per-request costs: search with contents at $7/1k requests, summaries at $1/1k, Exa Deep at $12/1k, and Deep Reasoning at $15/1k.
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.
Key Features
- Web search API with sub-150ms latency
- Token-efficient highlights (up to 90% reduction)
- Exa Agent for natural-language queries with effort modes
- Deep Agent for complex multi-step research
- Monitors for scheduled continuous data updates (webhook delivery)
- Structured output extraction (e.g., company data across 70M+ companies)
- Semantic search across code, docs, and web
- Supports auto, low, medium, high effort levels
- Highlights with maxCharacters control
- Zero data retention option available
- SOC 2 Type II certified
- Single Sign-On integration
- API playground for testing queries
- MCP Server support
- Websets for pattern-based search
Real-world workflow fit
Concrete scenarios for the personas Exa actually fits — and what changes day-one when you adopt it.
AI agent developer
You're building a coding agent that needs to search latest API docs and Stack Overflow answers.
Outcome: Integrate Exa Search API via Python SDK, set search type to 'auto' (~1s), use highlights for token-efficient content. Agent fetches relevant snippets with 90% token reduction, reducing LLM costs and latency.
Sales operations lead
You need to enrich a list of 100 target accounts with CEO names and founded dates for outreach.
Outcome: Use Exa's structured output with output_schema on company search. For $0.70 (100 requests at $7/1k), get clean JSON with CEO and founded_year. Feed directly into CRM via Zapier.
Product manager for a news monitoring tool
You want to track mentions of 'AI regulation' across news sites and get daily summaries.
Outcome: Set up Exa Monitors to run daily searches on News vertical, receive webhook alerts with AI-generated summaries ($15/1k requests). Integrate with Slack via Zapier for team updates.
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.
Models Under the Hood
Proprietary deep search modelHighlights extraction model
Limitations
Pricing is per-call and can stack up on agentic workloads that issue many searches per task — set rate limits and cache aggressively. Neural search quality depends on phrasing; keyword fallback exists. Index coverage on very fresh news (last few hours) and non-English content is thinner than Google. Free tier is limited to 1,000 requests/month. As of April 2026, the /research endpoint and several parameters (resolvedSearchType, highlightScores, startCrawlDate/endCrawlDate) are deprecated and will sunset May 1, 2026. Deep search modes take 12-40 seconds, which may be too slow for real-time chat applications.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Annual total
Free
Over 12 months
Effective monthly
Free
Billed monthly
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.
Free
$0/mo
Ideal for
Developers prototyping AI agents or testing Exa's API with up to 1,000 requests per month.
What this tier adds
Free entry point with all endpoints available; limited to 1,000 requests/month.
Search
$7/1k requests
Deep Search
$12-15/1k requests
Agent
$0.025-2.00/run
Contents
$1/1k pages
Monitors
$15/1k requests
Enterprise
Custom
Ideal for
Large-scale deployments requiring SOC 2 Type II, zero data retention, SSO, dedicated support, and custom rate limits.
What this tier adds
Adds security compliance, custom integrations, SLA, volume discounts, and dedicated onboarding.
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 pricing is moderate for production-grade APIs: Search at $7/1k requests is cheaper than Perplexity's specialized APIs but pricier than Brave Search ($5/1k). Deep Search and Agent tiers ($12-15/1k, $0.025-2/run) are premium for advanced use. The free tier (1k requests/month) lets you prototype. Enterprise gets volume discounts. For startups, the $1000 credits program helps. Overall, Exa fits teams that need accuracy and structured outputs over cost optimization.
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.
API key and basic integration: 5-10 minutes using the dashboard onboarding that generates idiomatic code for your stack (Python, JS, etc.). For custom workflows with deep search or monitors, budget 1-2 hours to configure parameters and test. Enterprise onboarding includes 1:1 support.
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.
Migrating in
- →From Bing Search API: Replace Bing calls with Exa's /search endpoint; map page content extraction to highlights for token efficiency.
- →From Perplexity API: Switch to Exa's /search with output_schema for structured results; adjust latency profiles as needed.
- →From Brave Search API: Use Exa's Python SDK and MCP server for agentic workflows; benefit from higher accuracy on FRAMES benchmark.
Migrating out
- ↗To Perplexity API: Convert Exa search calls to Perplexity's chat/completions endpoint; note lower accuracy on structured tasks.
- ↗To Brave Search API: Replace Exa's /search with Brave's /web endpoint; lose structured outputs and highlights for simpler scraping.
- ↗To Google Custom Search: Replace Exa's API with Google's JSON API; note higher costs and lack of agent-friendly highlights.
Recent material changes
Pricing, brand, ownership, or deprecation changes worth knowing before you commit. Most-recent first.
- •2026-04-01: Deprecated /research endpoint, resolvedSearchType, highlightScores, startCrawlDate/endCrawlDate — sunset May 1, 2026.
- •2026-03-01: Launched Exa Monitors for tracking web changes with webhook alerts.
- •2026-03-01: Updated pricing — Search at $7/1k, Deep Search $12-15/1k, new Agent pricing tiers.
- •2026-02-01: Introduced Exa Instant Search (180ms) for real-time applications.
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
Mcp Websets
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
- Documentationexa.ai
Openapi Spec
Full product docs from exa.ai
- Documentationexa.ai
Best Practices
Full product docs from exa.ai
- Documentationexa.ai
Migrating From Bing
Full product docs from exa.ai
- Documentationexa.ai
Building A Voice Agent With Exa
Full product docs from exa.ai
Frequently Asked Questions
Featured Head-to-Head Comparisons
Exa vs Firecrawl
For structured, token-efficient search with low latency and enterprise-grade features (SOC 2, SSO), Exa is the stronger choice, especially with its new Agent API and Deep Agent. For teams needing full web scraping, interaction, and an open-source, freemium starting point, Firecrawl offers unmatched flexibility with its keyless tier and broad integrations. Choose Exa for production AI agent search; choose Firecrawl for flexible data gathering and scraping at scale.
Exa vs Tavily
Tavily edges ahead for teams needing production-grade uptime (99.99% SLA), lower latency than Exa's 150ms, flexible pay-per-query via x402, and deep integrations (LangChain, MCP). Exa excels when structured outputs (e.g., company enrichment) and token-saving highlights matter, plus its new Exa Agent simplifies natural-language research. If SLA and raw speed are critical, choose Tavily; for structured data extraction and token efficiency, pick Exa.
Popular in Developer Infrastructure
Used Exa? Help shape our editorial sentiment research.