Apify Pricing: Real Cost per 1,000 Pages

Every scraping vendor has one number it does not publish. For the credit-based APIs it is the multiplier on a defended page. For Apify it is cost per page, and the reason it is missing is structural rather than evasive: Apify does not know. The platform rents you compute and a marketplace of other people's scrapers, and how many pages a gigabyte-hour buys is a property of the Actor, not of the platform.

That makes Apify hard to compare against a flat per-page API, and it makes the usual side-by-side table useless. So this is a teardown of the actual meter, with the arithmetic done out loud. Every figure below was read from apify.com/pricing and the Apify platform documentation in September 2026, not from a software directory, because directories in this category go stale within weeks of a repricing.

Apify pricing plans in 2026

Four tiers. Note the third column, because it is the unusual part.

The included credit equals the plan price on every tier. So the subscription is not a bundle at a discount, it is a prepayment with a volume discount attached to the compute rate. Moving from Starter to Business does not buy you cheaper pages so much as a 35% cheaper compute rate, far more RAM per run and eight times the concurrency. Annual billing takes 10% off the plan fee.

One thing to put in the calendar rather than the spreadsheet: unused credit does not roll over. It expires at the end of the billing period. If your crawling is seasonal, you are sizing the plan to your busiest month and donating the difference in the quiet ones.

How Apify actually meters usage

The unit is the compute unit, and the definition is refreshingly plain: one CU is one gigabyte of RAM running for one hour. An Actor configured with 4 GB that runs for thirty minutes consumes 2 CUs. At the Starter rate of $0.20 per CU, that run costs $0.40, whether it returned nine pages or ninety thousand.

Read that twice, because it inverts how you normally think about scraping spend. On a per-page API, an inefficient scraper costs you the same as an efficient one, and you only pay more if you ask for more pages. On Apify, an Actor that sleeps between requests, retries badly, or holds a browser open while it waits on a model, bills you for the waiting. Runtime is the product.

What a page really costs on Apify

Here is the arithmetic, done at the Starter rate of $0.20 per CU. These are derived figures, not published ones: they follow directly from the CU definition and the RAM and throughput assumptions in the first two columns, which you should replace with measurements from your own run.

Top row to bottom row is a factor of roughly 317, on one plan, at one published rate. That is not a criticism of Apify. It is the honest consequence of selling compute, and it is why anybody who tells you "Apify costs X per page" is quoting one Actor they happened to run.

The practical move is to stop guessing. Run your target Actor on 500 URLs, read the CU consumption off the run detail, divide, and you have your real number. Do it on the Free tier with the $5 credit before you subscribe. It takes twenty minutes and it is the only figure that matters.

The Actor layer that sits on top

Platform compute is not the whole bill when you rent from the Store. Apify supports a pay per usage model, where "users can run the Actor without any additional charges beyond the platform usage costs generated by the Actor," and a pay per event model, where the developer defines billable events in the Actor's source code, which can include "generating a single result or starting an Actor." Creators choose whether to pass platform costs through. There is also a rental model with a monthly fee for the Actor itself. On every paid plan those Actor charges are deducted from your prepaid credit; on the Free tier rented Actors are trial only.

We did not find a firsthand figure for Apify's platform commission on paid Actors, so we are not going to quote one. The buyer consequence stands without it: a Store Actor can carry a per-result price, a per-event price or a flat monthly rental, on top of compute, and two Actors that scrape the same site can price completely differently. Read the pricing tab on the Actor, not just the plan page.

Proxies, storage and the add-ons

Metered separately, and easy to miss when you are budgeting from the plan price alone. Residential proxy runs $8 per GB falling to $7 on higher plans. Datacenter proxies include a pool of IPs with overage from $1 down to $0.60 per IP. SERP queries are $2.50 per 1,000 falling to $1.70. The Unblocker is $1.50 per 1,000 requests falling to $1. Extra concurrent runs are $5 each and extra Actor RAM is $1 per GB. Personal training, if you want it, is $150 an hour.

Residential bandwidth is the line that surprises people. At $8 per GB, a crawl averaging 500 KB of transfer a page costs $4 per 1,000 pages in proxy alone, which on most of the table above dwarfs the compute. If your targets need residential IPs, price the bandwidth first and the plan second.

How does Apify pricing work?

You subscribe to a plan that credits your account with the same dollar amount you paid, then spend it on compute at a per-CU rate, where one CU is a gigabyte of RAM running for an hour. Rented Actor fees, proxy bandwidth, SERP queries and storage draw from the same credit. Unused credit expires monthly.

Does Apify have a free plan?

Yes, permanently. The Free tier costs $0 and includes $5 of platform credit, 16 GB maximum RAM, five concurrent runs and community support. Rented Actors are trial only on it. It is genuinely enough to benchmark an Actor and derive your own cost per page before you commit, which is the best use of it.

Do Apify credits roll over to the next month?

No. Unused usage credits are not rolled over and expire at the end of the billing cycle. That makes the plan a floor on your spend rather than a budget, so size it to your typical month and pay overage in the spikes rather than the other way around. A bill that swings several times over month to month is also a genuine annoyance for whoever has to reconcile the recurring vendor invoices at your end.

Is Apify cheaper than a flat per-page scraping API?

It depends entirely on the Actor, which is the honest answer and also the unsatisfying one. If your target is a plain server-rendered site and you have a lean HTTP Actor, nothing with a flat per-page rate will get close to $0.014 per 1,000 pages. If your target needs a browser and the Actor holds one open for five or ten seconds a page, flat per-page pricing wins by a wide margin and gets more predictable at the same time.

For reference on the other side of that line, our own web scraping API pricing is $39 a month for about 50,000 pages, $99 for about 250,000 and $399 for about 1.5 million, which works out at $0.78, $0.40 and $0.27 per 1,000 pages, with JavaScript rendering included and no surcharge for schema-typed JSON. The number does not move when a page is slow. That is the trade: you give up the marketplace and the browser automation, and you get a figure you can put in a forecast. We have laid the feature differences out in detail on our Apify alternative page, including the places Apify is clearly the better buy.

When Apify is the right purchase

Three cases, and they are real. When somebody has already written and maintained the exact Actor you need for a site you would otherwise reverse engineer yourself, renting it is cheaper than a sprint. When the job involves logging in, filling forms or clicking through a flow, Apify does browser automation and a crawl API does not. And when you need a full platform, with schedules, storage, webhooks, proxy management and a UI your non-engineers can use, that is what you are buying and it is worth money.

The case against is narrower than the internet suggests. It is not that Apify is expensive; at the top of that table it is the cheapest thing in the category. It is that the cost is a property of code you did not write, so you cannot forecast it from a price list, and finance teams hate that more than they hate a higher number.

If the conclusion you reach is that you would rather own the crawler outright than rent either compute or pages, that is a legitimate third road, and it has its own bill. We costed it out in detail for the leading open source option in Firecrawl self hosted, including the five services the Docker stack asks you to operate and the AGPL question that tends to stop commercial teams. For very large one-off jobs where predictability matters more than features, our bulk web scraping page has the volume math.

The decision, in one line

Do not compare Apify on its price list, because the price list does not contain the number you need. Run your candidate Actor on 500 URLs on the free tier, read the compute units off the run, and derive your own cost per 1,000 pages. Then compare that against a flat rate, and pick predictability or the marketplace deliberately rather than by accident.