Find AI Tools
Exa screenshot

What is Exa?

Exa is a web search engine and API designed for AI agents and applications rather than human browsing. It returns relevant pages, extracted highlights, full page contents and structured JSON outputs through a single API, with configurable latency from fast keyword-style queries to deep multi-step research. The platform also offers Websets for building structured datasets from the web, and Monitors for tracking fresh events on a schedule.

Key Features

Search API

Web search tool calls for agents with configurable latency from roughly 180ms to 1 second, returning relevant URLs and metadata.

Contents extraction

Returns full page text and token-efficient AI highlights that can cut token usage by up to 90 percent for LLM context.

Structured outputs

Returns results in customisable JSON schemas, with coverage cited at over 70 million companies and category filters for news, companies, research, people and financials.

Deep and Deep-Reasoning Search

Multi-step agent workflows that produce answers with web-grounded citations and structured outputs.

Agent runs

Asynchronous research runs with selectable effort modes (low to x-high) that return structured outputs with citations.

Monitors

Scheduled search runs that identify fresh events on the web and push webhook updates.

Enterprise security

SOC 2 Type II certification, Zero Data Retention options, Single Sign-On and custom rate limits.

Pros & Cons

Advantages

  • Pricing is transparent and usage based, with a free tier of up to 1,000 requests per month so developers can test before paying.
  • A single API covers search, content extraction, structured outputs and deeper research, reducing the need to stitch together multiple services.
  • Token-efficient highlights help keep LLM context costs down, which matters for high-volume agent workloads.
  • It is purpose-built for AI agents and cites strong results on retrieval benchmarks such as FRAMES, Tip-of-Tongue and Seal0.
  • Enterprise options include SOC 2 Type II, Zero Data Retention and SSO for teams with compliance requirements.

Limitations

  • The pricing model is granular, with separate charges for search, contents, deep search, monitors and agent runs, which can make total cost hard to predict.
  • It is an API-first developer tool, so non-technical users will need engineering help to integrate it.
  • Higher-volume needs, custom datasets and SLAs require contacting sales for custom Enterprise pricing.
  • Deep and agent modes add latency (deep search around 10 seconds, agent runs longer) compared with the fast search option.

Use Cases

Developers building AI agents that need live web search and grounded citations instead of relying on a model's training data.

Teams building RAG pipelines that want full page contents and token-efficient highlights for LLM context.

Companies generating structured datasets of firms, people or financials using Websets and structured JSON outputs.

Researchers and analysts running multi-step deep research queries that return cited, structured answers.

Product teams adding scheduled Monitors to track fresh news or events and trigger webhook notifications.

Enterprises needing search with SOC 2 Type II, Zero Data Retention and SSO for regulated environments.