EF
Productivity

Exa (formerly Metaphor)

Web search API for AI agents, with cited results and structured outputs.

At a glance

Free + paid · Source review 2026-10-26
Price & billing

Starter: free, with $10 in credits every month plus a $10 onboarding bonus and no payment method required. Developer: pay as you go. Enterprise: custom. Endpoint prices: Lite search from $4 per 1,000 requests, Auto search $7, Deep search $12, Contents $1 per 1,000 pages, Monitors $15 per 1,000 requests, Answer from $5. Currency is not explicitly labeled.

Free allowance

Starter: free, with $10 in credits every month plus a $10 onboarding bonus and no payment method required.

Best suited to

Engineering teams building coding agents, research copilots, RAG pipelines or sales-enrichment flows that need programmatic, citation-ready web access with clean excerpts.

Fit is editorial guidance, not a measured result.
Platforms

Access: API. Operating systems not separately verified.

Published-source facts and editorial guidance. No product test is implied. Confirm current conditions before purchase.

No product screenshot is claimed for this listing. Use the official website ↗ to inspect the current interface.

What to know before choosing

Good fit

Engineering teams building coding agents, research copilots, RAG pipelines or sales-enrichment flows that need programmatic, citation-ready web access with clean excerpts.

Limits and trade-offs

Pricing is per request and endpoint, and usage is metered by calls, pages and agent runs; check the endpoint table for the call type you use. Exa is a search API for agents and applications, not a consumer chat product.

Plans and free allowances

Starter: free, with $10 in credits every month plus a $10 onboarding bonus and no payment method required. Developer: pay as you go. Enterprise: custom. Endpoint prices: Lite search from $4 per 1,000 requests, Auto search $7, Deep search $12, Contents $1 per 1,000 pages, Monitors $15 per 1,000 requests, Answer from $5. Currency is not explicitly labeled.

Evidence

Official Exa homepage (formerly Metaphor) ↗ (reviewed 2026-10-26), Official Exa endpoint pricing ↗ (reviewed 2026-10-26), Former Metaphor homepage, redirects to exa.ai ↗ (reviewed 2026-10-26)

Official Exa pages captured and reviewed on Oracle; former Metaphor identity confirmed by redirect; no authenticated product test

Dates: Last reviewed 2026-10-26. Last reviewed is the latest source or editorial review date; it is not a content-change date.

A closer look at Exa (formerly Metaphor)

Exa (formerly Metaphor) is a search API built for AI agents. Instead of ten blue links, it offers neural, embedding-based retrieval with endpoints for search, crawling, answers and page contents, returning clean structured results for LLM pipelines.

Editorial guidance based on published documentation. The workflow below is a suggested starting point, not a measured product result.

When to choose Exa (formerly Metaphor)

  • Engineering teams building coding agents, research copilots, RAG pipelines or sales-enrichment flows that need programmatic, citation-ready web access with clean excerpts.
  • Teams that want the contents endpoint for full page text or deep search for multi-step research in the same API.

When to choose something else

  • Not for end users — there is no consumer search app here. A poor fit for casual browsing.
  • For teams that just need cheap keyword SERP data, where legacy SERP APIs cost less per thousand calls.

A practical starting workflow

  1. Get an API key (Starter is free with $10 in monthly credits plus a $10 onboarding bonus; no payment method required).
  2. Call the search endpoint with a natural query and retrieve embedding-ranked results with highlights.
  3. Use the contents endpoint for full page text, or deep search for multi-step research.
  4. Scale to a paid plan when credits run out.

Check before committing

  • Pricing is per request and endpoint: Lite search from $4 per 1,000 requests, Auto search $7, Deep search $12, Contents $1 per 1,000 pages, Monitors $15 per 1,000 requests, Answer from $5. There is no Core plan on the current pricing page; confirm the endpoint table on exa.ai/pricing.
  • Costs scale fast at agent volumes — model your query mix before committing. Endpoint pricing changes periodically; confirm on exa.ai/pricing.

Common questions

Is Metaphor the same as Exa?

Yes. Metaphor rebranded to Exa in 2024; metaphor.systems now redirects to exa.ai, and existing integrations still resolve.

Former Metaphor homepage, redirects to exa.ai ↗

How much does the Exa API cost?

Starter is free with $10 in credits every month plus a $10 onboarding bonus and no payment method required. Developer is pay as you go: Lite search from $4 per 1,000 requests, Auto search $7, Deep search $12, Contents $1 per 1,000 pages, Monitors $15 per 1,000 requests, Answer from $5. Enterprise is custom. There is no Core plan on the current pricing page.

Official Exa endpoint pricing ↗

Should I use Exa or a SERP API?

Use Exa when you need semantic, LLM-ready retrieval with clean excerpts and citations for agents. Use a SERP API (Serper, Brave) when you need cheap raw keyword results at high volume.

Official Exa homepage (formerly Metaphor) ↗

Compare approaches for this task

  • Tavily — Developers building AI agents and RAG pipelines who want search plus page extraction from one vendor, with framework integrations and a no-card free tier for prototyping.
  • Perplexity — Finding source leads for changing topics, then checking the original pages before drafting a comparison or brief.
  • Phind — Developers who want cited, up-to-date answers to coding and debugging questions without wading through search results, and who work in VS Code.
  • Glean — Enterprises that want one AI assistant and search layer across company applications.; Teams that need permission-aware access with observability over queries, answers and actions.

Sources and checks

Last source review is not a date of product testing. No rating or test result is implied by inclusion. Check the current plan and terms before buying.