Exa

Research

Exa is a neural, embeddings-based web search API built for AI agents, returning semantically relevant results plus clean page content. It is purpose-built neural search optimized for AI agents rather than keyword matching, with combined search-and-content retrieval and a structured-data product.

How to use Exa

Call the search API with natural-language or neural queries to get semantically ranked results, and add the contents option to retrieve clean page text, billed separately. Use the agentic and deep-search tiers for multi-step agent work, and the structured-data product to compile collections. Start free with a monthly search allowance plus starter credits.

Best for

Semantic web search for retrieval and AI agents. Clean full-page content extraction alongside search. Building structured datasets from search results.

How it behaves

An API-first product with no consumer UI, priced per request in layers where search and content extraction are billed separately and higher-reasoning tiers cost more.

Tips

Use neural queries for meaning, not just keywords. Add content extraction only when you need page text, it is billed separately. Reach for deep-search tiers for agent workflows.

Pricing

Standard search is about $7 per thousand requests, content extraction adds a dollar per thousand, and deeper tiers cost more. A free tier gives a monthly search allowance plus starter credits.