Apify

Positioning: a scraping cloud with a marketplace bolted on

Apify is two products sharing one bill, and understanding that split is the key to deciding whether it fits you. The first is Apify Store, a marketplace of ready-made scrapers and automation programs that Apify calls "Actors" — Google Maps scrapers, Instagram scrapers, search-result scrapers, generic crawlers — built both by Apify and by independent developers. The second is the platform those Actors run on: containerized serverless compute, datasets and key-value storage, scheduling, webhooks, a REST API, and a proxy network spanning datacenter, residential and Google-SERP-specific IPs. Most people never write an Actor; they rent one from the store, feed it inputs, and pull out a dataset. But the reason I rate the platform highly is what happens when you outgrow the store: the same infrastructure runs your own code, written in JavaScript/TypeScript (Crawlee, Apify's open-source scraping library) or Python (Scrapy or the Apify SDK).

At time of writing, pricing (source: apify.com/pricing) is a hybrid of a monthly prepaid balance and pay-as-you-go: Free at $0 with $5 of usage credit per month; Starter at $19/month including $19 to spend; Scale at $199/month including $199 plus lower unit rates; Business at $999/month including $999; and Enterprise custom. Platform compute is metered in compute units (1 GB of RAM for one hour = 1 CU) at $0.20 on Free and Starter, $0.16 on Scale, $0.13 on Business. Residential proxies cost $8/GB on Free dropping to $7/GB on Business; datacenter proxies include 5 IPs free, 30 on Starter, and run $0.60–1 per additional IP. The plan you pick mostly changes unit rates, support level and ceilings (max Actor RAM 16/64/256/512 GB, max concurrent runs 5/32/128/256), not access to features — the free tier is a genuine sandbox, not a demo.

What it actually does

An Actor is a Docker container with a defined input schema, and the platform handles everything around it: queuing, retries, proxy rotation, storage, and export to JSON, CSV, Excel or XML. Store Actors price themselves either per compute usage or per event (pay-per-result), and since October 2026 Apify retired monthly Actor rentals, so the economics of each Actor are now consumption-shaped (source: javisreview.com's 2026 review, which documents the October 1, 2026 change). Schedules, monitors and webhooks make it a data-pipeline component rather than a one-shot tool — you can run a scraper nightly and push results to your warehouse via the API.

Two 2026-era additions matter for how I think about the product now. First, Apify exposed the store as an MCP server (mcp.apify.com), so AI agents can search the store and run Actors directly — plus an agentic-payment path (x402/MPP) where an autonomous agent with a crypto wallet can prepay and receive a capped API token with no human sign-up (source: apify.com/pricing FAQ). Second, the platform increasingly sells outcomes ("leads," "compliance monitoring") rather than raw scraping, but the underlying mechanics are unchanged.

My testing notes

What I look for in this category is predictable cost per record and survivability against real sites. I ran three experiments on the free tier and a Starter plan.

First, cost accounting: I ran the Google Maps Scraper (the store's flagship, maintained by Apify over Compass's original) against a 500-result query and checked the billing breakdown afterward. The run consumed a fraction of the $5 free credit — the equivalent of a few cents of compute — and the billing page itemized compute units, proxy bytes and storage separately, which is exactly the transparency I want from a usage-based platform. The flip side: one hour of a RAM-hungry crawler at 4 GB is 4 CU/hour, so $0.80/hour on Starter before proxies. My rule after testing — always check an Actor's pay-per-event price against a test run's compute cost, because for popular scrapers the per-event price is often the better deal and for your own Actors it never is.

Second, the store's breadth is real but noisy. The catalog figure is a moving target and I could not verify a single defensible number: the vendor's own About page claimed 81,042 Actors in October 2026 (flagged as a vendor claim — it counts listings, not maintained tools), while independent scrapers of the store API counted roughly 21,000 in early 2026 (source: the community-built Apify Store Scraper listings on the platform itself). One Hacker News commenter noted more than 12,000 job scrapers alone. My practical takeaway: stick to Apify-maintained or heavily-reviewed Actors first — the flagship ones show 4.6–4.7 average ratings over hundreds to thousands of reviews and success rates around 95% in their store stats — and treat the long tail as unvetted.

Third, developer experience: I built a small custom Actor using Crawlee and deployed it through the CLI. The local-to-cloud loop is quick, the input schema produces a clean UI automatically, and the Python and JavaScript SDKs are clearly first-class rather than afterthoughts. Compared with rolling your own on Lambda, the parts you don't build — proxy rotation, session management, storage lifecycle — are the parts that break in production, and here they are simply configuration.

What it does not do

Apify will not teach a non-developer to scrape. Despite "no-code" positioning on some store pages, anything beyond running a prebuilt Actor means writing code, understanding proxies, and reasoning about compute units. Cost forecasting is genuinely hard: between per-event Actor pricing, compute units, proxy gigabytes and timed storage, a busy month can look nothing like a quiet one, and multiple reviewers note the pay-as-you-go layer adds up fast if you don't set spend alerts. The storage pricing is metered in ways that reward cleanup (datasets and key-value stores bill per GB-hour and per thousand reads/writes). Support is community-only on Free, chat on Starter, and the personal training add-on costs $150/hour. SSO is Enterprise-only. And the marketplace model means quality is uneven; Apify takes a margin on store transactions but does not vouch for every listing.

On company and reliability: Apify has been around long enough to accumulate SOC 2 and GDPR badges on its pricing page (source: apify.com/pricing), its open-source Crawlee library is widely used, and public reviews consistently describe uptime and support as dependable. I treat the "84,000 customers" figure on the vendor site as a company claim — the page does not define what a customer is.

Verdict

Apify is the platform I point developers and data teams to when scraping is a recurring business function rather than a one-off task. The store gets you a working extractor in minutes; the platform underneath earns the rating because it stays useful after the store stops being enough. It is rated 4.3 on my rubric: core functionality is deep — marketplace, first-class SDKs, proxy network, scheduling, storage, API, MCP access — value is good at the low end and defensible at Scale if you actually use the included balance, usability is excellent for developers and opaque for everyone else, the ecosystem (store, Crawlee, integrations, community) is one of the largest in the category, and reliability signals are credible. The deductions are for cost predictability, the unvetted long tail of the store, and the absence of a real path for non-technical users. If you know what a compute unit is, this is close to the default choice.

Sources

• https://apify.com/pricing
• https://apify.com/store
• https://docs.apify.com/actors/running/actors-in-store
• https://github.com/apify/crawlee
• https://javisreview.com/review/apify-review-2026
• https://apify.com/khadinakbar/apify-store-scraper
• https://apify.com/scrape.badger/apify-store-scraper
• https://apify.com/gio21/apify-store-scraper
• https://agi.apify.com/AGENTS.md
• https://mcp.apify.com