Exa logo

Exa

Web search API that gives AI agents clean, cited results

About Exa

## Intro Exa is a search engine built for AI rather than people. Instead of returning a page of blue links, its API returns results with page text and highlights that an AI agent can read directly. Builders use it to give chatbots, research agents and enrichment workflows access to fresh information from the web, with sources attached. Search speed is configurable, from a few hundred milliseconds for quick lookups up to around a second for higher quality results, so you can tune it for a chat interface or a background job. ## Key features - **Search.** Fast web search tuned for agents, returning text and highlights. You can trade speed for depth, from instant search to deep reasoning search. - **Contents.** Pull the full, cleaned text of any page for use as model context, with options for how fresh the crawl should be. - **Agent and Deep Search.** Hand Exa a research question and it runs a multi-step search, returning a structured answer with citations. Useful for company research or competitor scans. - **Monitors.** Schedule recurring searches that watch for new events on the web and send updates to a webhook, which plugs neatly into tools like n8n or Make. - **Answer and Websets.** An Answer endpoint returns a direct, cited answer to a question, and Websets helps build lists of companies or people that match criteria you describe. The range of endpoints is Exa's main advantage. You can start with simple search, then add page contents, research runs or monitoring as your product grows, all on one account and one bill. The trade-off is that pricing differs by endpoint, so it takes a little planning to estimate costs. ## Best for Exa suits builders adding live web knowledge to an AI product: a support bot that checks docs, a sales tool that researches companies, or a newsletter workflow that finds fresh stories. It also fits automation users who want reliable search results inside a workflow without scraping Google. You do need to work with an API or a platform that has an Exa integration. If you only want to search the web yourself, a consumer AI search tool is a better fit. ## Pricing summary The free plan includes $10 in credits each month plus a $10 onboarding bonus, with no card required. After that it is pay as you go: search runs from $4 to $15 per 1,000 requests depending on depth, and page contents cost $1 per 1,000 pages. Agent research runs are priced per request by effort level. Enterprise plans are custom. ## Quick verdict Exa is one of the better search APIs for AI agents, with more depth options than most rivals and a useful monitoring feature. The pay-per-request model is cheap for prototypes but needs watching once traffic grows, especially on deep search and agent runs. If your product needs current web data with citations, start on the free credits and benchmark it against Tavily.