Exa review, pricing and verdict

Last updated: 2026-09-04

Exa is a neural web search API built for AI agents and RAG pipelines, encoding pages as vector embeddings instead of matching keywords. Exa Instant responds in under 200 milliseconds, and every result returns clean extracted text rather than raw HTML, cutting the token overhead for downstream LLM calls.

About Exa

Exa is an AI-native web search API built by Exa Labs (San Francisco, founded 2021) for large language models, autonomous agents, and retrieval-augmented generation pipelines. Harvard classmates Will Bryk (CEO) and Jeffrey Wang started the company roughly a year before ChatGPT launched, betting that AI needed a search engine built for machine consumption rather than human browsing.

The core mechanism is embeddings-based retrieval: Exa encodes every indexed page as a dense vector using its own trained neural networks, then runs nearest-neighbour lookups when a query arrives, surfacing pages whose vectors sit closest to the query. That captures conceptual proximity keyword search misses. Every request returns clean extracted text rather than raw HTML, cutting the token overhead of the downstream LLM call.

There are three account plans. Starter is free and needs no payment method: Exa gives $20 in credits on sign-up plus $10 in credits every month, which it markets as "over $120 in credits per year". It carries 10 Search QPS, 50 agent concurrency, access to every endpoint, MCP server access, a Claude Connector and a ChatGPT plugin, and all four indexes — web, people, company and scholarly works. Developer is pay-as-you-go with no commitment, adding usage-based billing, email support, payment by card or bank, and up to 25 Search QPS. Enterprise adds a custom MSA and DPA, zero data retention, HIPAA alongside SOC 2 Type II, SSO and SCIM, monthly invoicing, SLAs, a dedicated Slack channel, custom QPS and agent concurrency, up to 1,000 results per search, custom indexes and bring-your-own data provider.

Usage is priced per product rather than per seat. Search is $7 per 1,000 requests with configurable latency from 180ms to 1s; Answer is $5 per 1,000; Deep Search is $12 per 1,000 for deep-lite and deep, $15 per 1,000 for deep-reasoning; Monitors are $15 per 1,000; Contents is $1 per 1,000 pages per content type, and results beyond the first ten cost another $1 per 1,000. Agent runs are billed per run across five fixed effort modes from $0.012 (minimal) to $1.00 (xhigh), or metered at $0.10 per Agent Compute Unit with search tool calls at $0.005 and email and phone enrichment at $0.02 and $0.07.

Websets, Exa's curated-dataset product, is priced separately as a subscription: Starter at $49/month for 8,000 credits and 100 results per Webset, Pro at $449/month for 100,000 credits and 1,000 results per Webset, and a custom Enterprise tier reaching 5,000 results per Webset.

One thing to read before sending sensitive queries: Exa's privacy policy states that "Query Data is used to improve our products and technology, including by training and fine-tuning models". Exa also says it does not monitor queries for personal information and advises against submitting personal data through query fields. Zero data retention is available, but it is an Enterprise plan feature rather than a default. Exa publishes a Vanta trust centre carrying SOC 2 Type II, a HIPAA attestation dated 17 June 2026, a data processing addendum, a penetration testing report and a subprocessor list, all behind access request.

Screenshots

Pricing

Two separate pricing systems. The API is usage-priced with no subscription: Search $7 per 1,000 requests (plus $1 per 1,000 for results beyond the first ten), Answer $5 per 1,000, Deep Search $12 per 1,000 for deep-lite and deep and $15 per 1,000 for deep-reasoning, Monitors $15 per 1,000, and Contents $1 per 1,000 pages per content type — so a search that also pulls full page text runs about $8 per 1,000. 07 per phone enrichment.

The free Starter plan needs no payment method and gives $20 in credits on sign-up plus $10 in credits every month; Developer is the same pricing with pay-as-you-go billing and up to 25 Search QPS instead of 10. Websets is a separate subscription: $49/month for 8,000 credits (100 results per Webset, 2 team seats) or $449/month for 100,000 credits (1,000 results per Webset, 10 team seats), with a custom Enterprise tier. Enterprise API pricing is custom and is where zero data retention, HIPAA, SSO/SCIM and custom QPS live.

Key Features

• Neural semantic search: Every indexed page is encoded as a vector embedding, so a query is matched by meaning rather than keyword overlap. Exa runs its own index rather than reselling Google or Bing results, and search latency is configurable from 180ms to 1s depending on how much depth the caller wants.
• Four indexes, not one web crawl: Web, people, company and scholarly-works indexes are all available on the free Starter plan. Agents query a pre-filtered vertical instead of sifting generic open-web results, and Enterprise can add custom tailored indexes or bring its own data provider.
• Contents: clean text instead of HTML: Returns parsed page text, query-relevant highlights and AI summaries at $1 per 1,000 pages per content type, removing the scraping layer and the markup that would otherwise consume LLM context. Livecrawl policies are configurable per request.
• Agent: per-run deep research: Asynchronous multi-step research runs with structured outputs and citations, billed per run across five effort modes from $0.012 to $1.00, or metered by Agent Compute Unit. Free and Developer plans both allow 50 concurrent agents.
• MCP server, Claude Connector and ChatGPT plugin: A free Model Context Protocol server plus a Claude Connector, a ChatGPT plugin and 50+ integrations, all included on the free tier, so Claude, Cursor, VS Code and similar tools reach Exa search without custom API glue.
• Monitors and Exa Connect: Monitors ($15 per 1,000) run saved searches on a schedule and surface what is new, and Exa Connect routes requests to third-party data providers — self-serve on Developer, premium providers on Enterprise.

Pros

• The free Starter plan is genuinely usable for evaluation and needs no payment method: $20 in credits on sign-up plus $10 every month, with access to every endpoint, all four indexes, 10 Search QPS and 50 agent concurrency rather than a crippled preview
• Ships a free official MCP server, a Claude Connector and a ChatGPT plugin, so Claude, Cursor and VS Code reach Exa without a custom REST wrapper — and those integrations are on the free tier, not gated behind a paid plan
• Pricing is per-request and published per product, so cost can be modelled before writing code: $7 per 1,000 searches, $1 per 1,000 pages of content, and a per-run agent price you pick by effort level rather than discover after the fact
• Runs its own neural index rather than reselling a third-party search engine, and offers people, company and scholarly-works indexes alongside the open web for knowledge-intensive agents

Cons

• Exa's privacy policy states that "Query Data is used to improve our products and technology, including by training and fine-tuning models". Zero data retention exists but is an Enterprise plan feature, so on Starter and Developer the queries an agent sends are training input by default
• Content extraction is billed on top of search — $1 per 1,000 pages per content type, and another $1 per 1,000 for results past the first ten — so a search that returns full text and a summary costs materially more than the headline $7 per 1,000
• Default rate limit is 10 Search QPS. Pay-as-you-go raises it to 25, but anything beyond that needs an Enterprise conversation, so a high-concurrency pipeline still has to build queuing and retry logic or talk to sales
• It is an API, not a product: there is no interface for a non-technical team. The one no-code surface, Websets, is a separate $49-to-$449/month subscription rather than part of the API plan

Visit Exa Official Website