Exa

No ratings yet

Semantic web search API for AI agents - 400K+ developers use meaning-based retrieval; $357M raised at $2.2B valuation; Series C closed May 2026.

Exa (formerly Metaphor Systems, rebranded January 2024) is a semantic web search API built for AI agents and LLM applications rather than human end users. It uses transformer-based neural embeddings to search the web by meaning - not keywords - across 500B+ indexed URLs built on proprietary crawlers and vector databases. Unlike Google or Bing APIs, Exa returns conceptually related results for LLM grounding, with token-efficient highlights that reduce context consumption by up to 90%. Exa raised a $250M Series C led by a16z at a $2.2B valuation in May 2026, bringing total funding to $357M. Customers include Cursor, Cognition (Devin), HubSpot, AWS, and Databricks. A free tier of 1,000 requests/month is available with no card required.

#ai-search

#search-api

#developer-tools

#llm

#research

#semantic-search

Freemium

Free plan available

View Pricing

Update Tool

exa.ai

Freemium

Pricing Model

Research & Analysis

Key Features

Neural semantic search - retrieves web content by meaning rather than keywords; query as a sentence, paragraph, or URL to find conceptually matching pages
Token-efficient highlights - extracts the most relevant excerpts from pages, reducing LLM context consumption by up to 90% vs. ingesting full pages
Structured JSON company and contact extraction from 70M+ company profiles for lead enrichment and sales intelligence pipelines
Variable latency modes - Fast (~180ms), Auto (~1s), Deep (~10s) for different accuracy/speed trade-offs within the same API
Monitors - continuously track specific web topics or signals and deliver results via webhook for event-driven AI agent pipelines
Specialized verticals: company data, people search, code, financial filings, and news for domain-specific agent applications
Full page content retrieval for feeding entire articles and documents into LLM context without building custom scrapers

Use Cases

AI agent developers grounding LLMs in live web content without building custom scrapers or managing rate limits across multiple search APIs
Research agents using semantic search to find conceptually related sources that keyword matching would miss entirely
Startups building sales intelligence tools using Exa's company and people search for lead enrichment at $0.02/email and $0.07/phone
Data scientists feeding real-time market signals and financial filings into LLM-powered research pipelines via the finance and news verticals

Pros

400K+ developers and $357M raised at $2.2B valuation (Series C, May 2026) - the most funded and most-adopted semantic search API in the AI developer ecosystem
Token-efficient highlights reduce LLM context by 90% - dramatically lowers the per-query cost of building retrieval-augmented generation applications
Customers include Cursor, Devin, HubSpot, and Databricks - validated by the highest-quality AI-native product teams in the industry

Cons

Pure pay-per-use pricing after the free tier - costs are unpredictable for agents making dynamic or unbounded search calls without built-in rate limiting
Free tier of 1,000 requests/month is consumed in hours by a production agent with moderate search frequency - meaningful testing requires a paid setup quickly
Developer-only API with no consumer interface - there is no way to test or explore search results without writing code, which raises the evaluation barrier