Exa

Overview

Exa is a search engine built for machines rather than people. Founded in 2021 by Will Bryk and Jeff Wang as Metaphor Systems and renamed Exa in January 2024, the company runs its own crawler and index and represents pages as embeddings, so queries are matched on meaning instead of keyword overlap. It went through Y Combinator, raised $85 million at a $700 million valuation in September 2025, and followed with a $250 million Series C led by Andreessen Horowitz in May 2026 at a reported $2.2 billion valuation — roughly $361 million raised in total.

The practical appeal for agent builders is that Exa returns cleaned page content with the results. A conventional search API hands back links, leaving you to fetch, strip and chunk each page yourself; Exa’s Contents endpoint returns text, highlights or summaries directly, which removes a whole scraping layer. Beyond plain search, the product line has expanded into Deep Search for multi-step research, an Answer endpoint that returns cited answers, Monitors for scheduled searches that watch for new events, and Websets for building enriched lists of companies or people.

Pricing is genuinely usage-based — no subscription, $7 per 1,000 searches, and free monthly credits — but the meter has many dials. Extra results, additional content types, AI summaries, agent compute units and contact enrichment each bill on their own line, so real-world cost per query is often well above the headline rate. Index breadth is also narrower than Google’s, especially outside English. Exa is the right choice when semantic recall and content-in-the-response matter more than exhaustive coverage.

Key Benefits

• Semantic recall: Descriptive queries (“startups doing X in Europe”) return relevant pages that keyword search misses.
• Content included: Full text, highlights and summaries come back with results, cutting the scrape-and-parse step out of the pipeline.
• Right-sized endpoints: Use cheap search for simple lookups and deep-research or agent endpoints only where the task warrants it.
• Frictionless start: No subscription, free monthly credits and MCP support mean you can wire it into an agent in minutes.

Use Cases

1. Retrieval for AI agents — Give an agent live, content-rich web access without building crawling and parsing infrastructure.
2. Automated research reports — Chain Deep Search and the Answer endpoint to produce cited briefs on a topic.
3. Prospect and market list building — Use Websets to assemble and enrich lists of companies, people or research papers matching precise criteria.
4. Competitive and news monitoring — Schedule Monitors to surface new pages about a company, product or topic as they appear.