Instances

List instances

runcrate instances list
runcrate instances list -q    # IDs only
runcrate ps                   # Alias for instances list

Browse GPU types and pricing

runcrate instances types
runcrate instances types --gpu A100
runcrate instances types --gpu H100 --region us-east

Deploy an instance

# Auto-pick a config for a GPU type: the best-priced one that's starting reliably
runcrate instances create --name training-run --gpu A100

# Deploy with a specific template
runcrate instances create --name training-run --gpu A100 --template ubuntu-train

# Deploy with a storage volume (mounts at /workspace)
runcrate instances create --name training-run --gpu A100 --volume my-volume

# Deploy a specific config from `instances types`
runcrate instances create --name training-run --type-id <id>

Stop an instance automatically

# Terminate 4 hours after it starts running
runcrate instances create --name training-run --gpu H100 --max-runtime 4h

# Terminate once the GPU sits under 5% with nobody connected for 30 minutes
runcrate instances create --name notebook --gpu A100 --idle-stop 30m

# Both: whichever comes first
runcrate instances create --name job --gpu H100 --max-runtime 1h30m --idle-stop 30m

Flag	Format	Range
`--max-runtime`	Duration with a unit, whole minutes: `30m`, `90m`, `1h30m`, `4h`, `168h`	`5m` to `168h` (7 days)
`--idle-stop`	Duration with a unit, whole minutes: `15m`, `30m`, `2h`	`5m` to `24h`

Both are off when omitted. Days (7d), bare numbers (30) and seconds (30s) are rejected before anything launches. Either way the instance is terminated and its local disk deleted. If the environment has reached its monthly budget, create fails with “This environment has reached its monthly budget. An owner or admin can raise it in the dashboard.” See Spend guardrails.

Change auto stop on a running instance

# Extend before a long job: max runtime counts from when it started, so give the total
runcrate instances auto-stop training-run --max-runtime 8h

# Add an idle stop, or turn a limit off
runcrate instances auto-stop notebook --idle-stop 30m
runcrate instances auto-stop job --max-runtime off

Pass --max-runtime, --idle-stop or both; off turns one off and a flag you leave out keeps its value. A change never stops the instance sooner than 2 minutes from now. It prints the new limits and the auto stop time.

Instance details and status

runcrate instances info <name-or-id>      # Full details (IP, GPU, cost, guardrails)
runcrate instances status <name-or-id>    # Live status check

instances info also prints:

Row	Shows
`Max runtime:`	The limit, such as `4h` or `1h30m`, or `off`
`Idle stop:`	The idle window, such as `30m`, or `off`
`Auto stop at:`	When max runtime will terminate it (only when set)
`Stop reason:`	Why it stopped, once it has: Stopped on request, Reached its max runtime, Stopped when idle, Environment budget reached, Balance ran out, Failed, or Stopped by Runcrate

With --json, the raw max_runtime_minutes, idle_shutdown_minutes, auto_stop_at and stop_reason fields are included.

Terminate an instance

runcrate instances delete <name-or-id>

Welcome

SDKs

CLI

Inference

Compute

Clusters

Jobs

Schedules

Secrets

Storage

Dedicated Clusters

Billing

Account

Webhooks

Integrations

Hosts

FAQ

List instances

Browse GPU types and pricing

Deploy an instance

Stop an instance automatically

Change auto stop on a running instance

Instance details and status

Terminate an instance

​List instances

​Browse GPU types and pricing

​Deploy an instance

​Stop an instance automatically

​Change auto stop on a running instance

​Instance details and status

​Terminate an instance

List instances

Browse GPU types and pricing

Deploy an instance

Stop an instance automatically

Change auto stop on a running instance

Instance details and status

Terminate an instance