Exa
The search engine built for AI — power your own sourcing agents
4.4Usage-based; free tierBest for Funds with technical DNA building proprietary sourcing and research agents
Search API Sourcing Agents AI Infrastructure
Overview
Exa (formerly Metaphor) is a meaning-based search engine built for AI applications: an embeddings-powered web index queried by semantic intent rather than keywords, exposed through an API plus Websets, which turns searches like 'seed-stage devtools companies with ex-FAANG founders' into structured, enriched lists. Y Combinator-alum and venture-backed, it has become a core primitive for funds building their own sourcing and research agents.
Key features
- Semantic web search API that matches meaning, not keywords
- Websets: searches returned as structured lists of companies and people
- Similarity search — find companies like a given URL
- Full-page content retrieval for downstream AI processing
- Enrichment columns (funding, headcount signals, contact paths)
Pros
- Purpose-built substrate for custom deal-sourcing agents
- Finds long-tail companies keyword search and databases miss
- Developer-friendly, with transparent usage pricing
Cons
- An API-first tool — non-technical teams should start with Websets only
- Coverage depends on the public web; stealth companies stay invisible
- Entity data is thinner than dedicated company-data vendors