M
About
Exa, formerly known as Metaphor AI, is an AI-native web search API and website crawler designed to supercharge AI agents, RAG pipelines, and LLM applications with real-time, high-quality web data. Unlike traditional search APIs, Exa is purpose-built for machine consumption — returning structured outputs, relevant excerpts (highlights), and crawled content that reduce token usage and improve agent performance. Exa operates dedicated, high-quality web indexes for specific verticals including people search, company data (70M+ companies), code documentation, financial data, and news — outperforming competitors like Perplexity and Brave on FRAMES, Tip-of-Tongue, and Seal0 benchmarks. Its Exa Instant mode returns results in under 180ms, making it the fastest search provider available. Key capabilities include structured output extraction, token-efficient content highlights, deep web research agents (Exa Deep), and a full MCP Server for seamless integration with AI toolchains. Enterprise features include SOC 2 Type II certification, Zero Data Retention (ZDR) options, and Single Sign-On (SSO). Exa is trusted by leading AI companies including OpenRouter and is ideal for developers building AI agents, coding assistants, research tools, lead generation systems, and any application requiring grounded, up-to-date web intelligence.
Key Features
- Ultra-Low Latency Search: Exa Instant returns search results in under 180ms — faster than any other search provider — making it ideal for real-time AI agent applications.
- Structured Output Extraction: Automatically extracts structured JSON outputs from web content and Exa's database of 70M+ companies, enabling complex data enrichment workflows.
- Vertical-Specific Web Indexes: Dedicated high-quality indexes for people, companies, code documentation, financial data, and news deliver superior accuracy over general-purpose search.
- Token-Efficient Content Highlights: Highlights extract only the most relevant excerpts from web pages for a given query, reducing LLM token usage and costs by over 50%.
- Enterprise-Grade Security: SOC 2 Type II certified with Zero Data Retention options and Single Sign-On support for teams requiring compliance and privacy controls.
Use Cases
- Powering AI agents and LLM applications with real-time, grounded web search results
- Building company and people research pipelines using structured data from 70M+ company records
- Augmenting RAG (Retrieval-Augmented Generation) systems with fresh, relevant web content
- Reducing LLM inference costs by extracting token-efficient page highlights instead of full content
- Integrating live web intelligence into coding assistants for up-to-date documentation retrieval
Pros
- Best-in-Class Accuracy: Outperforms competitors on FRAMES, Tip-of-Tongue, and Seal0 — the most demanding retrieval benchmarks — making it the most reliable choice for AI grounding.
- Purpose-Built for AI Agents: Structured outputs, highlights, and MCP Server integration make Exa a natural fit for LLM pipelines and agentic workflows without extra parsing overhead.
- Free Tier Available: Developers can try the API for free and evaluate performance before committing to a paid plan, lowering the barrier to adoption.
- Specialized Vertical Search: Dedicated indexes for people, companies, and code outperform generic search APIs in domain-specific retrieval tasks.
Cons
- Developer-Only Access: Exa requires API integration and technical knowledge — there is no consumer-facing no-code interface for non-developers.
- Cost at Scale: High-volume search workloads can become expensive, and enterprise pricing requires contacting the sales team for custom agreements.
- Web Search Scope Only: Exa is focused exclusively on web search and retrieval; it does not offer generative AI, summarization, or chat interfaces natively.
Frequently Asked Questions
What is Exa, and how is it different from Metaphor AI?
Exa is the rebranded version of Metaphor AI. It is a web search API and AI search engine purpose-built for developers and AI agents, offering real-time web data retrieval with structured outputs, highlights, and sub-180ms latency.
How fast is Exa's search API?
Exa Instant returns search results in under 180ms, which the company claims is faster than any other search provider currently available.
What verticals does Exa support beyond general web search?
Exa maintains dedicated indexes for people search, company data (70M+ companies), code documentation, financial data, and news — each optimized for higher accuracy in those specific domains.
Can I try Exa for free?
Yes, Exa offers a free tier so developers can test the API and evaluate its performance. A paid plan is required for higher usage volumes and enterprise features.
Does Exa support structured data extraction?
Yes. Exa can extract structured JSON outputs directly from web pages and its company database, enabling automated enrichment workflows without additional parsing steps.
