Exa, formerly Metaphor, is a search engine and API designed specifically for AI applications and developers. Rather than relying purely on keyword matching, Exa uses neural search to retrieve semantically relevant content and returns rich, structured results suited to retrieval-augmented generation, research agents, and competitive intelligence workflows.
Exa's core idea is that AI agents need a search layer built for machines rather than humans. It indexes and processes large portions of the web and exposes endpoints that let developers give AI products real-time, current knowledge, powering thousands of AI workflows that require fresh information beyond a model's training cutoff.
The company was founded in 2021 by William Bryk and Jeffrey Wang, who built a search engine from scratch focused on being both fast and high quality for AI consumption. Exa went through Y Combinator and has built its product around developer-first APIs.
In September 2025 Exa raised an $85 million Series B led by Benchmark at a $700 million valuation, with participation from Lightspeed, Y Combinator, and NVIDIA's venture arm NVentures. Benchmark's Peter Fenton joined the board. The capital is earmarked for substantially expanding compute, scaling indexing to capture more of the web, and hiring across engineering and go-to-market.
Exa differentiates through neural retrieval quality, speed optimized for agentic use, and a developer-centric API surface. It competes with both general search APIs and AI-native search products, positioning itself as infrastructure for builders rather than a consumer search destination.