Exa is a search API built for AI: embeddings-based retrieval over the web, with content extraction and an answer endpoint on top. It is a neural search engine sold by the request.
It is the only competitor in this family with no monthly plan at all, and that is the most interesting thing about its pricing.
What it costs
No subscription, no minimum spend. Per-request rates:
| Product | Rate |
|---|---|
| Search (instant) | $4 / 1,000 |
| Search (fast/auto) | $7 / 1,000 |
| Deep Search | $12–15 / 1,000 |
| Contents | $1 / 1,000 pages |
| Answer | $5 / 1,000 |
| Monitors | $15 / 1,000 |
| Agent | $0.012–$1.00 per request |
Free tier: $10 in credits — about 2,500 instant searches — and it resets to $10 on the first of every month. No payment method required; completing an account adds a one-time $10 bonus.
Why no-commitment matters more than the rate
$4 per 1,000 searches is expensive next to a raw-fetch credit — about 53× the $0.075 of ScrapingBee. Unit definition again: Exa searches, ranks and extracts; a scraper fetches one URL you chose.
The structural difference is the commitment. Every other competitor here makes you pick a tier and pay monthly whether you use it or not — €300 at Data365, $149 at Ayrshare, $49 at Oxylabs. Exa charges what you used and nothing when you do not.
For an unproven product that is a real advantage, and it is the one thing on this page we cannot match: our free tier is 1,000 credits a month, and past that you buy credit packs.
⚠️ The free tier resets, which is rarer than it sounds
"Your free balance resets to $10 on the first of every month."
Compare Xpoz, whose 500 free credits are a one-time allocation that does not renew. A renewing free tier is a materially different offer, and the two are routinely listed side by side as "free tier" in comparison tables — including, until this page, in ours.
Who it's for
An agent that needs to search and read the open web: research, question answering, retrieval where the corpus is the internet. The Contents endpoint at $1 per 1,000 pages is genuinely cheap for ingestion.
It does not answer what gained traction in this niche this week. No search API does, because that needs readings taken hourly and kept, and the past cannot be photographed retroactively.
Where it is better than us
No commitment. Pay for what you use, nothing when idle. We cannot match it.
The renewing free tier. $10/month in credits, forever, no card.
The open web. Eleven platforms against everything.
Neural retrieval. Ranking by meaning is a harder problem than ours and they solve it.
The legal part
A search API sits on crawled public content; the compliance surface belongs to the crawl. We are not going to characterise their position from outside.
Ours is published: public versus personal data and a removal route.
Who measured this
We did, and we barely compete — for web search we are not an alternative, and a page implying otherwise would be the kind of comparison this family exists to avoid.
Seven rates, the free-tier amount and the monthly reset are all read from their page. The 53× comparison is our arithmetic, and it is the number we would most warn you against using, because it divides two different units.
What we run is an hourly archive of what's rising across public platforms. What we sell is one response shape across eleven platforms — a different question from the one Exa answers well.
All seven per-request rates, the $10 free-credit amount, the ~2,500-instant-search figure and the monthly reset read from exa.ai/pricing on 2 October 2026 — primary source, read directly from the page. Exa publishes no monthly subscription tiers; the absence is theirs, not an omission of ours. The 53× comparison is our arithmetic across two different units and is flagged as such. Verify before deciding.