Web search built for AI agents

Exa provides a single API for search, crawling, and research agents, with configurable latency from 180ms to 1 second. It supports real-time search data with token-efficient page contents, making it suitable for coding agents, chatbots, and voice agents.

Agents get fast, accurate web search with reduced token usage.

Token-efficient content extraction

Exa's specialized model extracts the most relevant excerpts from web pages, reducing token consumption by up to 90%. This is powered by a model trained to identify key information, enabling efficient context for AI models.

Lower token costs and improved context relevance for AI applications.

Structured data extraction and enrichment

Exa can extract enrichments as structured outputs across 70M+ companies, including fields like company name, CEO, and founding year. This is useful for lead generation, company research, and data enrichment tasks.

Automated, structured data collection for business intelligence.

Deep research with citations

Exa offers Deep Search and Deep-Reasoning Search endpoints that perform multi-step agent workflows, returning answers with structured outputs and web-grounded citations. This supports complex queries and research tasks.

Comprehensive, cited answers for in-depth research.

Automated web monitoring

Exa Monitors identify fresh events on the web by running searches at a specified cadence (e.g., daily or weekly) and sending updates via webhooks. This enables real-time tracking of new information.

Stay updated with automated alerts on new web content.

Enterprise-grade security and controls

Exa offers Zero Data Retention (ZDR) for privacy compliance, SOC 2 Type II certification, and Single Sign-On (SSO) for team authentication. These features ensure secure and compliant usage for enterprises.

Secure, compliant search infrastructure for enterprise teams.

• High accuracy across demanding benchmarksExa leads across FRAMES, Tip-of-Tongue, and Seal0 — the most demanding retrieval benchmarks — with accuracy scores of 54.4%, 54.2%, and 36% respectively, outperforming Perplexity and Brave in each category.
• Low-latency search optionsExa Instant returns results in under 180ms, faster than any other search provider. Configurable latency ranges from 180ms to 1s, making it suitable for latency-sensitive agent loops.
• Token-efficient content with highlightsUp to 25T tokens are passed to models each week as concise excerpts powered by highlights. A specialized model extracts the most relevant excerpts from the web, reducing token usage by up to 90%.
• Structured outputs for enrichmentExa extracts enrichments as structured outputs across 70M+ companies. The API can return data like company name, CEO name, and founded year in a structured JSON format.

• Pricing varies by endpoint and usageSearch costs $7 per 1k requests, Deep Search $12 per 1k requests, Deep-Reasoning Search $15 per 1k requests, Contents $1 per 1k pages, Monitors $15 per 1k requests, and Answer $5 per 1k requests. Additional results above 10 cost $1 per 1k requests for Search, Deep Search, and Deep-Reasoning Search.
• Agent pricing is usage-basedAgent uses effort: auto by default, with compute and tool usage scaling to the task. Fixed effort modes range from $0.012 per request (Minimal) to $1.00 per request (X-high). Contact enrichment is billed separately: $0.02 per email and $0.07 per phone number.
• Enterprise features require contacting salesEnterprise plans offer up to 1,000 results per search, requests with more than 25 results, custom rate limits (QPS), tailored moderation, custom index, enterprise-grade support, and SLAs. Pricing is not listed and requires talking to the sales team.
• Free tier provides limited creditsNew users get $20 credits on sign-up, and the free tier includes $10 credits per month. This allows limited usage of web search tool calls, webpage text and highlights, and configurable latency options.

AI Agent Web Search

Exa provides a single API for search, crawling, and research agents, designed to make agents smarter with token-efficient search across docs and repos. It powers coding agents with up to 90% token reduction, and supports configurable latency from 180ms to 1s for fast agent loops. Trusted by teams like Cognition to power Devin.

Company and People Enrichment

Exa enables structured outputs across 70M+ companies, extracting enrichments like CEO names, founding years, and financial data. It powers agentic lead engines for HubSpot and supports people search and company search verticals, with best-in-class accuracy for these specialized queries.

Deep Research and Structured Outputs

Exa's Deep Search runs multi-step agent workflows with answers and structured outputs, grounded in web citations. It offers async agents for deep research, list building, and enrichment, with pricing from $0.012 to $1.00 per run depending on effort mode. This is ideal for complex queries requiring comprehensive, cited answers.

Real-Time News and Event Monitoring

Exa Monitors identify fresh events on the web by running searches at a specified cadence (daily or weekly) and receive updates via webhooks. This is useful for news monitoring, tracking company updates, and staying current with changelogs. Pricing is $15 per 1k requests.

Sign up on Exa.ai and create your first workspace.

Connect your document storage (Google Drive, SharePoint, or S3).

Run the auto‑indexing wizard to ingest files.

Test a natural‑language query in the web console or Slack bot.