Why a web search API can cost this little
A price 18× under the nearest own-index competitor invites one obvious question: what is the catch, and how long does this last? Here is the answer, and here is what we will not tell you.
Five claims you can hold us to
What we will not tell you, and why
An earlier version of this page explained how the index is built and served in enough detail to reproduce it. That was a mistake, and we have taken it down. Building an index that stays this cheap to run is the hard part of this company, and there is no version of publishing it that helps you and does not help the competitors holding two billion dollars.
What that costs you is the ability to audit the price from first principles, and we would rather say so plainly than pretend the page was always this shape. What it does not cost you is anything you can act on. Every figure that touches your side of the API is still published and still checkable:
Measured latency
Published per mode with p50, p95 and p99, and with the measurement conditions attached rather than buried. See the numbers.
What is in the index
why_not(url) answers present, removed with a typed reason, or never admitted. No competitor does this.
Whether it works for you
10,000 requests a month, no card. Run your own evaluation set — that answers the question an architecture diagram never could.
What this page is not claiming
- That our index is as large as Google's, or as large as the $2 billion competitors'. It is not. It is bounded on purpose, and we will tell you what is missing.
- That cheaper is automatically better. If you need Google's exact ranking for SEO work, a SERP reseller is the right tool and we are not.
- That any of this is audited. These are our own claims about our own business, offered so the price has a stated explanation rather than an implied one. Weigh them as such.
Frequently asked questions
Is the price a loss-leader?
No. Margin is positive on every paid plan at current volumes, and it widens as volume grows rather than narrowing. We are not buying the market with the intention of repricing it later.
Why does margin improve with volume rather than shrink?
Because almost none of what we spend scales with query count. Our standing costs are close to fixed, and an additional query against content we already hold costs very little to serve. That is the opposite of a reseller, whose largest cost line is a per-call fee that grows in lockstep with its own revenue.
Will the price rise once you have customers?
We are self-funded and not raising, so there is no valuation the pricing has to grow into. That is the structural answer. The honest caveat is that we are also smaller than the companies we compete with, and you should weigh both.
Why not publish the full unit economics?
We used to. How the index is built and served is the part of this business that is actually hard, and writing the mechanism out in public was a mistake we have corrected. What you can still check is everything that affects you: the price, the rate limits, the measured latency, what is in the index and — through why_not — what is not.
How do I verify any of this without the internals?
By using it. The free tier is 10,000 requests a month with no card, which is enough to run your own evaluation set against your own queries and measure latency, coverage and result quality directly. A published architecture would not tell you whether it works for your workload; a month of real queries will.
What happens if a query misses the index?
You get an honest answer rather than a padded one, and the gap is recorded so the crawler can close it. why_not(url) will tell you whether a specific URL is present, was removed, or was never admitted.
$0.40 per 1,000 queries
Start on the free tier and check the latency, the coverage and the graph against your own workload.