Exa is a web search API and AI search engine designed specifically for AI agents, developers, and enterprise applications. It provides high-quality, low-latency search results through a neural retrieval approach that differs fundamentally from traditional keyword-based search engines. The platform indexes and serves content across hundreds of billions of webpages and offers dedicated indexes optimized for people search, company search, code documentation, financial data, and news.

The core product suite includes Search, which returns lists of results and page contents with latency profiles ranging from under 180ms (Instant) to standard speed; Contents, which delivers token-efficient webpage content for use in LLM pipelines; Answer, which provides direct responses backed by citations; and Agentic Search (Deep mode), which supports structured outputs and multi-step reasoning workflows. Websets is an additional product that allows users to build structured datasets from web search results.

Exa leads accuracy benchmarks on FRAMES, Tip-of-Tongue, and Seal0 — three of the most demanding retrieval evaluation sets — and outperforms Perplexity and Brave across these verticals. The platform processes queries in under 180ms for its fastest tier and returns structured outputs drawn from a database of over 70 million companies. Token-efficient highlights reduce LLM context costs by over 50% according to customers.

Enterprise features include Zero Data Retention, SOC 2 Type II certification, Single Sign-On, custom rate limits, SLAs, MSAs, and 1:1 onboarding support. Usage-based pricing starts at $1 per 1,000 page retrievals, with up to 1,000 free requests per month available on the free tier. Customers include Notion, Cursor, AWS, Databricks, HubSpot, Monday.com, Vercel, Lovable, Groq, OpenRouter, and others.