Apify features and the limits on each plan
Apify's feature set is one platform doing five jobs. A marketplace of prebuilt scrapers called Actors, a container runtime that executes them, three storage primitives, a proxy pool, and a scheduling and webhook layer that moves the output somewhere useful. Every one of those works on the Free plan. What money buys is ceiling, not capability. This page carries the ceilings, each re-fetched on 2026-09-09.
What Apify is not: a proxy vendor. Datacenter IPs come bundled, residential bandwidth is metered per gigabyte on every tier including Free, and that single line decides whether the platform is cheap or expensive for you.
What changes when you move up a plan?
Every figure below comes from apify.com/pricing and docs.apify.com/platform/limits, both fetched 2026-09-09.
| Limit | Free | Starter | Scale | Business |
|---|---|---|---|---|
| Monthly price | $0 | $19 ($17 annual) | $199 ($179 annual) | $999 ($899 annual) |
| Included platform usage | $5 | $19 | $199 | $999 |
| Combined memory, all running jobs | 16,384 MB | 65,536 MB | 262,144 MB | 524,288 MB |
| Memory for a single run | 16,384 MB | 32,768 MB | Not published | Not published |
| Concurrent Actor runs | 5 or 25 (sources conflict, see below) | 32 | 128 | 256 |
| Datacenter proxy IPs | 5 included | 30 included, then $1/IP | 200 included, then $0.80/IP | 500 included, then $0.60/IP |
| Residential proxy | $8/GB | $8/GB | $7.50/GB | $7/GB |
| Apify Store discount | None | Bronze | Silver | Gold |
| Support | Community | Chat | Priority chat | Account manager |
Add-ons are sold per unit, not as bundles. Concurrent runs cost $5 per run, Actor RAM $1 per GB, datacenter proxy from $0.60 per IP, and personal training $150 an hour (apify.com/pricing, 2026-09-09).
Three figures worth stating precisely: apify.com/store says 70,000+ Actors; paid Actor RAM runs 64/256/512 GB by plan; and priority chat support is bundled into Scale, with no separate add-on on the current pricing page.
Create a free Apify account if you want to test these ceilings against your own workload. The obvious objection first. That $5 monthly credit is use-it-or-lose-it, and Store Actors that charge per event bill on top of it. A free account can still produce a bill. The pricing breakdown and cost optimization guide cover how that happens.
Why is the RAM figure not the RAM you get?
The pricing page labels its memory row "Max Actor RAM", which reads like the size of the biggest run you can start. It is not. docs.apify.com/platform/limits publishes the identical numbers under a different label: "Maximum combined memory of all running jobs".
That page then lists a separate per-run cap. Free tops out at 16,384 MB for one run, Starter at 32,768 MB. So Starter's 64 GB is two 32 GB runs side by side, never one 64 GB run.
The distinction bites in one direction only. If your work is many small crawls, the combined figure is the one that constrains you. If your work is a single memory-hungry headless browser session, the per-run cap is the number to plan against, and Starter doubles Free rather than quadrupling it. Scale and Business publish no per-run maximum, so confirm yours in Console before you architect around a large single run.
How many Actors are actually in the Apify Store?
apify.com/store states "70,000+ ready-to-run tools called Actors" (checked 2026-09-09).
The public Store API is worth querying yourself, because the total depends on how you ask. api.apify.com/v2/store?limit=1&sortBy=newest returned 69,860 on 2026-09-09; the same endpoint with default sorting returned 57,157. Both are reproducible in one curl call.
Neither number is a useful buying signal. Catalog size tells you nothing about whether a maintained Actor exists for the one site you need, and Actors get deprecated without the count moving. Search the Store for your target, then check the Actor's last build date and run count before you commit a schedule to it.
Is Apify Proxy included, or metered?
Both, and the split matters more than any other line on the pricing page.
Datacenter IPs are included: 5 on Free, 30 on Starter, 200 on Scale, 500 on Business, with overage from $0.60 to $1 per IP. Residential bandwidth is never included. It bills at $8/GB subject to change · verified 2026-09-09 on Free and Starter, $7.50/GB on Scale, and $7/GB on Business, drawn against your platform credit.
Sites that only need datacenter IPs run essentially free on proxy. Sites that demand residential traffic do not. If you pull tens of gigabytes a month through residential IPs, price that bandwidth against a dedicated proxy vendor first. At $8/GB it dominates the bill long before compute does. Our guide to scraping dynamic websites covers how to tell which class of IP a target actually requires.
How long does Apify keep your data?
Named storages are "always exempt from deletion regardless of retention periods" (docs.apify.com/platform/storage, 2026-09-09). Unnamed storages are the ones that vanish.
On Free, your 10 most recent runs are retained for four months, and unnamed storages beyond those 10 are deleted when the retention period expires. Paid plans follow a retention window you configure in billing settings.
The practical rule is short. If a dataset matters past the current sprint, name it, or push it out to Sheets, S3, or a warehouse on the same run. The storage guide covers datasets, key-value stores, and request queues in detail, and compute units covers what the runs themselves cost.
What connects to Apify?
docs.apify.com/platform/integrations lists 25 entries across five sections (counted from the live page, 2026-09-09), not the "30+" this page previously claimed without a source. Two of those entries, Actor-to-Actor and integration-ready Actors, are platform mechanisms rather than outside services. The remaining 23 third-party integrations group into four families:
| Family | Named integrations |
|---|---|
| Programming | API, webhooks, GitHub |
| AI and agents | Apify MCP server, LangChain, OpenAI Agents SDK, Vercel AI SDK, Google ADK, Pinecone, Cursor, Claude Code CLI, Codex CLI, OpenCode, GitHub Copilot, Kimi Code CLI |
| Data and storage | Snowflake, Airtable, Google Drive, HubSpot |
| Workflows and notifications | n8n, Make, Zapier, Gumloop |
One correction, because older writing still gets it wrong. Apify's MCP server is a hosted endpoint at https://mcp.apify.com, not an Actor you run. Authentication is OAuth or a bearer token, and a plain request with no token returns 401 (checked 2026-09-09). Anonymous access is limited to discovery and documentation tools. A local @apify/actors-mcp-server npm package still exists for stdio transport during development. Our walkthrough of Apify MCP servers covers client setup, and the integrations hub covers the rest.
Where do Apify's own sources disagree?
On one number, and it is a number you plan schedules around.
Free-plan concurrency is listed as 5 concurrent runs on apify.com/pricing and as 25 concurrent runs on docs.apify.com/platform/limits. Both are Apify's own pages, both were fetched 2026-09-09, and neither references the other.
Assume 5 when you design a schedule. Being wrong in that direction costs you elapsed time. Being wrong the other way silently queues runs you expected to fire together. Then confirm the real ceiling on your account's Limits page in Console. The paid tiers agree across both sources, so 32, 128, and 256 are safe to plan against.
When is Apify the wrong tool?
Three situations where a smaller thing wins, stated plainly.
You already own a working crawler. If your scraper runs fine and the only thing failing is IP reputation, you need addresses rather than a runtime. Buying residential bandwidth from a dedicated proxy vendor avoids paying for compute, storage, and scheduling you will not use.
You need one extraction, once. A handful of pages pulled by hand does not justify an account, a Task, and a retention policy. A local script finishes faster than the platform onboarding does.
Your volume is residential-heavy and steady. At $8/GB on Free and Starter, a predictable monthly residential load is the case where a specialist proxy contract usually beats bundled bandwidth. Run that arithmetic before you upgrade a plan to fix a proxy problem.
Where Apify genuinely wins is the opposite shape: many different targets, changing often, where somebody else maintaining 70,000 scrapers is worth more than any single feature on this page.
Apify Store (70,000+ Actors per apify.com/store, checked 2026-09-09), the Crawlee library and language SDKs, a container runtime metered in compute units, three storage types (dataset, key-value store, request queue), Apify Proxy, Tasks and schedules, webhooks, a hosted MCP server at mcp.apify.com, and 23 third-party integrations documented at docs.apify.com/platform/integrations, which lists 25 entries across five sections including two Actor-to-Actor mechanisms (checked 2026-09-09).
Apify's own sources disagree. apify.com/pricing lists 5 concurrent runs on Free; docs.apify.com/platform/limits lists 25. Both were checked on 2026-09-09. Plan for 5 and confirm the real figure on your account's Limits page in Console. Paid tiers agree across both sources at 32 (Starter), 128 (Scale), and 256 (Business).
16,384 MB on Free and 32,768 MB on Starter, per docs.apify.com/platform/limits (2026-09-09). The larger figures quoted per plan (16 GB, 64 GB, 256 GB, 512 GB) are the maximum combined memory of all running jobs, not the size of one run. Scale and Business do not publish a per-run maximum.
Datacenter IPs are included at 5, 30, 200, and 500 per plan. Residential proxy is metered separately at $8/GB on Free and Starter, $7.50/GB on Scale, and $7/GB on Business (apify.com/pricing, 2026-09-09), billed against your platform credit.
Unnamed storages do expire: the 10 most recent runs are retained for four months on Free, and anything beyond that is deleted when the retention period ends. Named storages are exempt from deletion on every plan. Name a storage, or export the dataset in the same run, if you need the data later.
No. It is a hosted endpoint at https://mcp.apify.com that takes OAuth or a bearer token; an unauthenticated request returns 401. A local @apify/actors-mcp-server npm package exists for stdio transport in development, but the platform integration is the hosted URL.
Sources
Each fetched 2026-09-09.
- apify.com/pricing — plan prices, memory, concurrency, proxy allowances, add-ons, support tiers
- apify.com/store — Store size claim
- docs.apify.com/platform/limits — combined and per-run memory, concurrency
- docs.apify.com/platform/storage — retention for named and unnamed storages
- docs.apify.com/platform/integrations — the integrations catalog: 25 entries across five sections, 23 of them third-party
api.apify.com/v2/store— Store totals, queried directly