Apify Resources & Environment

Memory Allocation

Memory must be a power of 2, between 128MB and 32768MB (e.g., 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768 MB) [Usage and resources]
If not specified at run time, the Actor’s configured default memory is used
You can set minMemoryMbytes and maxMemoryMbytes in actor.json to constrain what users can allocate [Set memory limits]
Memory can also be dynamic — calculated from input before the run starts via a MathJS expression in actor.json [Dynamic memory]

CPU is automatically derived from memory: 1 full CPU core per 4096MB
- 512MB → 1/8 core, 1024MB → 1/4 core, 8192MB → 2 cores [Usage and resources]
On startup, an Actor may temporarily burst above its CPU allocation to start faster

Disk space = 2× allocated memory (e.g., 1024MB memory → 2048MB disk) [Usage and resources]

These are automatically injected into your container by the platform: [System env vars]

Variable	Description
`ACTOR_MEMORY_MBYTES`	Allocated memory in MB (use to derive CPU allocation)
`ACTOR_ID`	ID of the Actor definition
`ACTOR_RUN_ID`	ID of this specific run
`ACTOR_TASK_ID`	ID of the task (empty if run directly via API)
`ACTOR_DEFAULT_DATASET_ID`	Default dataset for output
`ACTOR_DEFAULT_KEY_VALUE_STORE_ID`	Default key-value store (input is stored here)
`ACTOR_DEFAULT_REQUEST_QUEUE_ID`	Default request queue
`ACTOR_INPUT_KEY`	Key in KV store where input is stored (usually `INPUT`)
`ACTOR_EVENTS_WEBSOCKET_URL`	WebSocket URL for receiving real-time platform events
`ACTOR_TIMEOUT_AT`	ISO 8601 timestamp when the run will time out
`ACTOR_STARTED_AT`	ISO 8601 timestamp when the run started
`ACTOR_WEB_SERVER_PORT`	TCP port for an optional HTTP server your container can expose
`ACTOR_WEB_SERVER_URL`	Public URL for the above server
`APIFY_TOKEN`	API token for calling Apify APIs
`APIFY_IS_AT_HOME`	`1` if running on Apify servers
`APIFY_PROXY_PASSWORD`	Password for Apify Proxy (if used)
`ACTOR_MAX_PAID_DATASET_ITEMS`	Max items for pay-per-result Actors
`ACTOR_MAX_TOTAL_CHARGE_USD`	Max charge for pay-per-event Actors

Connect to ACTOR_EVENTS_WEBSOCKET_URL to receive JSON messages: [System events]

{
  "name": "cpuInfo",
  "createdAt": "2024-01-01T00:00:00Z",
  "data": { "isCpuOverloaded": false }
}

Event	Payload	Use
`cpuInfo`	`{ isCpuOverloaded: bool }`	Detect CPU pressure, throttle concurrency
`migrating`	`{ timeRemainingSecs: float }`	Persist state before server migration
`aborting`	none	Clean up before graceful abort
`persistState`	`{ isMigrating: bool }`	Save state at regular intervals (every ~60s)

Global API rate limit: 250,000 requests/minute (per authenticated user) [API limits]
Per-resource rate limit: 60 requests/second per resource (some endpoints: 200–400 req/s)

Limit	Value
Max concurrent runs (Business plan)	256
Max log size	10,485,760 characters
Build timeout	1,800 seconds
Max metamorphs per run	10

CU = (Memory in MB / 1024) × (Duration in hours)

Example: 1024MB × 1 hour = 1 CU [Usage and resources]