Exa

What Exa does

Exa is an applied AI research lab that organizes human knowledge, starting with the web. It provides search APIs and web agents designed for AI, offering high-quality, token-efficient results with citations. Exa's products power various AI applications, including coding agents, deep research, list-building, and enrichment workflows. The company focuses on providing real-time web data and structured outputs to make AI agents smarter and more efficient.

Products

• Exa SearchA high-quality search API at every latency, accepting natural language queries and producing token-efficient results with citations. It offers various search types (auto, instant, fast, deep-lite, deep, deep-reasoning) and category-specific search for companies, people, publications, news, and personal sites. It also provides AI summary and structured outputs.
• Exa AgentA high-compute web agent that synthesizes information from across the web to handle complex deep research, list-building, and enrichment workflows. It offers structured outputs with web citations and various effort levels (minimal, low, medium, high, x-high, auto).
• Exa ConnectA platform to access leading data providers, offering full page web contents with highlights optimized for AI.
• Contents APIReturns clean, structured content from any URL, handling JavaScript-rendered pages, PDFs, and complex layouts. It can provide full page text, targeted highlights, or LLM-generated summaries, and can crawl linked subpages.
• Monitors APIRuns Exa searches on a recurring schedule and delivers results to a webhook endpoint, with automatic deduplication. Used to track competitor announcements, funding rounds, regulatory changes, and research publications.

Key capabilities

• Natural language queries
• Token-efficient results with citations
• Configurable latency (180ms to 1s for web search, up to 40s for deep-reasoning)
• Structured outputs
• AI summary
• Comprehensive coverage over all search verticals
• Specialized model for extracting relevant excerpts (highlights)
• Zero Data Retention (ZDR)
• SOC 2 Type II Certified
• Single-Sign On (SSO)
• Asynchronous deep research runs
• Access to leading data providers
• Full page web contents optimized for AI
• Configurable livecrawl policies
• Multi-step agent workflows
• Answers with web-grounded citations
• Daily/weekly monitoring of web events
• High-accuracy code retrieval for coding agents
• Search over millions of GitHub repos, docs, and Stack Overflow
• Sub-200ms search options for latency-sensitive agent loops
• Ground code generation with current package and API references
• Clean, structured content from any URL
• Handling of JavaScript-rendered pages, PDFs, and complex layouts
• Subpage crawling
• Content freshness control (cache or livecrawl)
• Scheduled search with automatic deduplication
• Webhook delivery for monitors
• Manual trigger for monitors

Use cases

• Powering coding agents with smart token-efficient search
• Deep research for AI agents
• List building for AI agents
• Enrichment workflows for AI agents
• Web search tool calls for agents
• Extracting enrichments as structured outputs across companies
• Integrating with AI agents for fast results or deep research
• Generating structured outputs from complex queries
• Tracking new events and updates across the web
• Searching docs, repos, and changelogs in real time
• Grounding code generation with current package and API references
• CRM enrichment
• GTM intelligence
• KYC/KYB insights
• Research reports
• Generating tables and filling missing data in Google Sheets
• Tracking funding announcements
• Monitoring competitor news with structured output
• Tracking research papers with highlights
• Prospecting, enrichment, qualification, and routing across CRM workspaces
• Identifying and reasoning over company and people data
• Automating complex web browsing and data extraction

AI approach

Exa is an applied AI research lab that builds a custom search engine for AI agents. It trains specialized models to extract relevant excerpts from the web, condense full webpages into token-efficient content, and process information to return structured outputs or grounded answers. Their products, Exa Search and Exa Agent, are designed to provide high-quality, low-latency web data for AI agents, including capabilities like deep research, list building, and enrichment.

Tech named: LLMs, model fusion, token-efficient highlights model

Industries served

• Software Development
• AI Engineering
• Fintech
• Biotechnology
• Sales
• Recruiting

What it says sets it apart

• Highest quality search at every latency
• Best-in-class accuracy across demanding retrieval benchmarks (FRAMES, Tip-of-Tongue, Seal0)
• Token-efficient contents with highlights (up to 90% token reduction)
• Specialized model trained to extract relevant excerpts
• Extracts enrichments as structured outputs across 70M+ companies
• Low-latency to deep research in one API
• Enterprise-grade security and controls (Zero Data Retention, SOC 2 Type II, SSO)
• Custom search engine built for AIs
• Custom indexes of 1B+ people, 50M+ companies, 350M+ publications
• Frontier quality at a fraction of the cost for web research
• Outperforms comparable systems across web research, entity enrichment, and financial analysis benchmarks
• Model fusion and token-efficient highlights model for lower cost and latency
• Pay-per-request with scalable effort levels
• Live web intelligence built for agents
• Structured output built for agents (typed fields directly into CRM records)
• Natural-language search over the live web (not stale databases)
• Continuous monitoring for real-time market changes
• Agentic scale for prospecting, enrichment, qualification, and routing
• Better than native model search experience for HubSpot

Funding rounds we track

From the AI funding tracker — rounds as reported by the linked publications.