spawn command reference

spawn launches, connects to, and manages ephemeral EC2 instances. This is the exhaustive command and flag reference. New to spawn? Start with Your first instance and the MPI/cohort guide.

Generated reference

This page is generated from the spawn CLI itself and kept in sync by a release-gated drift check, so every command and flag below matches the shipped binary. For task-oriented walkthroughs see the guides.

Global flags

These apply to every spawn command.

Flag	Short	Type	Default	Description
--accessibility		bool		Enable accessibility mode (implies --no-emoji)
--account		string		Expected AWS account ID (optional guard)
--lang		string		Language for output (en, es, fr, de, ja, pt)
--no-color		bool		Disable colorized output
--no-emoji		bool		Disable emoji in output
--output	-o	string	table	Output format (table, json)
--profile		string		AWS named profile (overrides SPORE_PROFILE/AWS_PROFILE and the shared config)
--region		string		Default AWS region (overrides SPORE_REGION/AWS_REGION and the shared config)
--verbose	-v	bool		Enable verbose output

spawn alerts

Create, list, and manage alert notifications for parameter sweeps and schedules.

Get notified via email, Slack, SNS, or webhooks when sweeps complete, fail, exceed cost thresholds, or encounter issues.

Examples:

Create alert for sweep completion

spawn alerts create <sweep-id> --on-complete --email user@example.com

Create alert for failures with Slack

spawn alerts create <sweep-id> --on-failure --slack https://hooks.slack.com/...

Create cost threshold alert

spawn alerts create <sweep-id> --cost-threshold 100 --email user@example.com

List all alerts

spawn alerts list

Delete alert

spawn alerts delete <alert-id>

spawn alerts

spawn alerts create

Create a new alert for a parameter sweep or schedule.

At least one trigger (--on-complete, --on-failure, etc.) and one destination (--email, --slack, --sns, --webhook) must be specified.

Examples:

Sweep completion via email

spawn alerts create sweep-123 --on-complete --email user@example.com

Multiple triggers and destinations

spawn alerts create sweep-123 \ --on-complete \ --on-failure \ --email user@example.com \ --slack https://hooks.slack.com/services/...

Cost threshold alert

spawn alerts create sweep-123 \ --cost-threshold 100 \ --email finance@example.com

Long-running sweep alert (trigger after 2 hours)

spawn alerts create sweep-123 \ --long-running 120 \ --email user@example.com

Schedule execution failure alert

spawn alerts create --schedule-id sched-123 \ --on-failure \ --slack https://hooks.slack.com/...

spawn alerts create <sweep-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--cost-threshold		float64		Alert when cost exceeds threshold (dollars)
--email		string		Email address for notifications
--instance-failed		bool		Alert when any instance fails
--long-running		int		Alert when sweep runs longer than N minutes
--on-complete		bool		Alert when sweep/schedule completes
--on-failure		bool		Alert when sweep/schedule fails
--schedule-id		string		Schedule ID (alternative to sweep-id)
--slack		string		Slack webhook URL for notifications
--sns		string		SNS topic ARN for notifications
--webhook		string		Webhook URL for notifications

spawn alerts delete

Delete an alert configuration.

Example: spawn alerts delete alert-abc123

spawn alerts delete <alert-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn alerts history

Show the history of notifications sent for an alert.

Example: spawn alerts history alert-abc123

spawn alerts history <alert-id>

spawn alerts list

List all alert configurations for the current user.

Optionally filter by sweep ID.

Examples:

List all alerts

spawn alerts list

List alerts for specific sweep

spawn alerts list --sweep-id sweep-123

spawn alerts list [flags]

Flags:

Flag	Short	Type	Default	Description
--sweep-id		string		Filter by sweep ID

spawn ami

Create and list AMIs created by spawn.

Examples: spawn ami list spawn ami list --stack pytorch --arch arm64 spawn ami create my-instance --name pytorch-2.4-cuda12

spawn ami

spawn ami create

Create an AMI from a running instance

spawn ami create <instance-id-or-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--description		string		Description for the AMI
--name		string		Name for the AMI (required)
--reboot		bool		Reboot instance before creating AMI (default: no-reboot)
--tag		stringArray		Tags in key=value format
--wait		bool		Wait for AMI to become available

spawn ami delete

Deregister a spawn-managed AMI and delete its backing EBS snapshots in one step. If the AMI was produced by EC2 Image Builder (e.g. 'spawn image import'), the corresponding Image Builder image resource is also deleted so its name/version is freed.

This is irreversible. Use 'spawn ami list' to find AMIs.

Examples: spawn ami delete ami-0123456789abcdef0 spawn ami delete ami-0123456789abcdef0 --region us-east-1 --yes

spawn ami delete <ami-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--region		string		AWS region (default: current region from AWS config)
--yes	-y	bool		Skip the confirmation prompt

spawn ami list

List spawn-managed AMIs

spawn ami list [flags]

Flags:

Flag	Short	Type	Default	Description
--arch		string		Filter by architecture (x86_64 or arm64)
--deprecated		bool		Show deprecated AMIs
--gpu		string		Filter by GPU support (true or false)
--region		string		AWS region (default: current region from AWS config)
--stack		string		Filter by stack (spawn:stack tag)
--version		string		Filter by version (spawn:version tag)

spawn ami snapshots

Show the EBS snapshots that back an AMI, with size, state, and whether each snapshot is shared with other AMIs (which is why 'spawn ami delete' keeps shared snapshots instead of deleting them).

Examples: spawn ami snapshots ami-0123456789abcdef0 spawn ami snapshots ami-0123456789abcdef0 -o json

spawn ami snapshots <ami-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--region		string		AWS region (default: current region from AWS config)

spawn app

Launch streamable research applications in the cloud.

Each application is pre-configured with the right instance type, NICE DCV for browser-based streaming, and automatic idle termination.

Apps you can launch depend on your account: public images are available to everyone; private images appear only if your account can pull them. Bring your own image with --image, or add bindings in ~/.spawn/catalog.yaml (--catalog).

Examples: spawn app list # show apps launchable from your account spawn app launch paraview # launch ParaView on a GPU instance spawn app launch igv --region us-west-2 spawn app launch paraview --spot --ttl 4h spawn app launch paraview --image 123456789012.dkr.ecr.us-east-1.amazonaws.com/paraview:5.13.2

spawn app

Flags:

Flag	Short	Type	Default	Description
--catalog		string		Local catalog overlay file (default: $SPAWN_CATALOG or ~/.spawn/catalog.yaml)

spawn app launch

Launch a catalog application via NICE DCV

spawn app launch <app-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--app-version		string		App container image tag to launch (default: catalog default; see 'spawn app list')
--idle-timeout		string		Stop when DCV has no clients for this duration (default: catalog default)
--image		string		Launch a BYO container image for this app (overrides the catalog binding), e.g. 123456789012.dkr.ecr.us-east-1.amazonaws.com/paraview:5.13.2
--instance-type		string		Override instance type (default: first catalog family + .xlarge)
--name		string		Session name (default: <app>-<timestamp>)
--no-open		bool		Write session file but do not open browser automatically
--region		string		AWS region (default: from AWS config)
--spot		bool		Use Spot pricing
--ttl		string		Hard termination deadline (e.g. 4h, 8h)

spawn app list

List all streamable applications in the catalog

spawn app list

spawn autoscale

Launch and manage auto-scaling job arrays that maintain target capacity

spawn autoscale

Flags:

Flag	Short	Type	Default	Description
--env		string	production	Environment (production or staging)
--table		string	spawn-autoscale-groups	DynamoDB table name

spawn autoscale health

Show instance health for auto-scaling group

spawn autoscale health <group-name>

spawn autoscale launch

Launch a new auto-scaling job array with specified capacity and launch template

spawn autoscale launch [flags]

Flags:

Flag	Short	Type	Default	Description
--ami		string		AMI ID (required)
--desired-capacity		int		Desired instance count (required)
--iam-profile		string		IAM instance profile
--instance-type		string		EC2 instance type (required)
--job-array-id		string		Job array ID (auto-generated if not specified)
--key-name		string		SSH key name
--max-capacity		int		Maximum instance count (default: desired * 2)
--metric-name		string		CloudWatch metric name (for custom metrics)
--metric-namespace		string		CloudWatch namespace (for custom metrics)
--metric-period		int	300	Metric evaluation period in seconds
--metric-policy		string		Metric policy type: 'cpu', 'memory', or 'custom'
--metric-statistic		string	Average	Metric statistic: Average, Maximum, or Minimum
--min-capacity		int		Minimum instance count (default: 0)
--name		string		Group name (required)
--queue-url		string		SQS queue URL for queue-depth policy (required if --scaling-policy=queue-depth)
--scale-down-cooldown		int	300	Scale-down cooldown in seconds
--scale-up-cooldown		int	60	Scale-up cooldown in seconds
--scaling-policy		string		Scaling policy type: 'queue-depth' (empty = manual mode)
--security-group-ids		stringSlice		Security group IDs (comma-separated or repeated)
--spot		bool		Use spot instances
--subnet-id		string		Subnet ID
--tag		stringArray		Additional tag key=value (repeatable)
--target-messages-per-instance		int	10	Target messages per instance for queue-depth scaling
--target-value		float64		Target metric value (e.g., 70.0 for 70% CPU)
--user-data		string		User data script (base64 encoded)

spawn autoscale list

Aliases: ls

List all active auto-scaling groups

spawn autoscale list

spawn autoscale metric-activity

Show recent metric-based scaling activity

spawn autoscale metric-activity <group-name>

spawn autoscale pause

Pause auto-scaling (stop reconciliation)

spawn autoscale pause <group-name>

spawn autoscale resume

Resume auto-scaling

spawn autoscale resume <group-name>

spawn autoscale scaling-activity

Show recent scaling activity for an autoscale group

spawn autoscale scaling-activity <group-name>

spawn autoscale schedule

Manage scheduled scaling actions for an autoscale group

spawn autoscale schedule

spawn autoscale schedule add

Add a scheduled action to an autoscale group

spawn autoscale schedule add <group-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--desired-capacity		int		Desired capacity (required)
--enabled		bool	true	Enable the schedule immediately
--max-capacity		int		Maximum capacity override (optional)
--min-capacity		int		Minimum capacity override (optional)
--name		string		Schedule name (required)
--schedule		string		Cron expression: 'second minute hour day month weekday' (required)
--timezone		string	UTC	Timezone (e.g., America/New_York)

spawn autoscale schedule list

List all scheduled actions for an autoscale group

spawn autoscale schedule list <group-name>

spawn autoscale schedule remove

Remove a scheduled action from an autoscale group

spawn autoscale schedule remove <group-name> <schedule-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn autoscale set-metric-policy

Set or update metric-based scaling policy

spawn autoscale set-metric-policy <group-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--metric-name		string		CloudWatch metric name (for custom metrics)
--metric-namespace		string		CloudWatch namespace (for custom metrics)
--metric-period		int	300	Metric evaluation period in seconds
--metric-policy		string		Metric policy type: 'cpu', 'memory', or 'custom'
--metric-statistic		string	Average	Metric statistic: Average, Maximum, or Minimum
--none		bool		Remove metric policy
--target-value		float64		Target metric value (e.g., 70.0 for 70% CPU)

spawn autoscale set-scaling-policy

Aliases: set-policy

Set or update scaling policy for an autoscale group

spawn autoscale set-scaling-policy <group-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--none		bool		Remove scaling policy (revert to manual mode)
--queue-url		string		SQS queue URL for single-queue policy (deprecated: use --queue)
--queue-weight		float64Slice		Queue weight 0.0-1.0 (must match number of --queue flags)
--queue		stringSlice		SQS queue URL (can be specified multiple times for multi-queue)
--scale-down-cooldown		int	300	Scale-down cooldown in seconds
--scale-up-cooldown		int	60	Scale-up cooldown in seconds
--scaling-policy		string		Scaling policy type: 'queue-depth'
--target-messages-per-instance		int	10	Target messages per instance for queue-depth scaling

spawn autoscale status

Show auto-scaling group status

spawn autoscale status [group-name]

spawn autoscale terminate

Terminate auto-scaling group and all instances

spawn autoscale terminate <group-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn autoscale update

Update auto-scaling group capacity

spawn autoscale update <group-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--desired-capacity		int	-1	New desired capacity
--max-capacity		int	-1	New maximum capacity
--min-capacity		int	-1	New minimum capacity

spawn availability

Display availability statistics based on historical launch success/failure data.

This helps identify regions with proven capacity for specific instance types. Statistics are passively collected from actual launch attempts.

spawn availability [flags]

Flags:

Flag	Short	Type	Default	Description
--instance-type		string		Instance type to check (required)
--regions	-r	stringSlice		Regions to check (comma-separated or repeated; default: common regions)

spawn burst

Launch EC2 instances that will register with the hybrid registry and coordinate with local instances to process workloads.

This enables "cloud bursting" where local compute capacity can be extended with on-demand cloud resources.

spawn burst [flags]

Examples:

# Launch 10 instances to help process a job array
  spawn burst --count 10 --job-array-id my-array --instance-type c5.4xlarge

  # Launch with specific AMI
  spawn burst --count 5 --job-array-id genomics --ami ami-abc123

  # Launch Spot instances for cost savings
  spawn burst --count 20 --job-array-id simulation --spot

Flags:

Flag	Short	Type	Default	Description
--ami		string		AMI ID (auto-detect if not specified)
--count		int	1	Number of instances to launch
--instance-type		string	t3.micro	EC2 instance type
--job-array-id		string		Job array ID to join (required)
--job-array-name		string		Job array name (optional)
--key-name		string		SSH key pair name
--security-group-ids		stringSlice		Security group IDs (comma-separated or repeated)
--spot		bool		Use Spot instances
--subnet-id		string		Subnet ID

spawn cancel

Deprecated: use 'spawn sweep cancel <sweep-id>' instead

Cancel a running parameter sweep and terminate all instances.

Queries DynamoDB for the sweep state, terminates all running/pending instances via cross-account access, and updates the sweep status to CANCELLED.

Examples:

Cancel a running sweep

spawn cancel --sweep-id sweep-20260116-abc123

spawn cancel [flags]

Flags:

Flag	Short	Type	Default	Description
--sweep-id		string		Sweep ID to cancel (required)
--yes	-y	bool		Skip the confirmation prompt

spawn capacity-block

Purchase and manage EC2 Capacity Blocks for ML.

Discover purchasable offerings with 'truffle capacity-blocks', then purchase one here. After purchase, launch into it with: spawn launch <name> --reservation-id <id> --capacity-block --az <reservation-az>

spawn capacity-block

spawn capacity-block purchase

Purchase a Capacity Block for ML from an offering id (from 'truffle capacity-blocks').

⚠️ A Capacity Block is billed UP FRONT and is NON-REFUNDABLE — the full block duration is charged at purchase. This is the single most expensive action spawn can take. The purchase requires you to TYPE three confirmations (the exact price, 'purchase <offering-id>', and an acknowledgement phrase) and refuses to run on a non-interactive terminal. Use --dry-run first to preview the price and terms without buying anything.

The offering's instance type, count, and duration must be supplied so the exact offering can be re-validated (and its current price re-confirmed) immediately before purchase.

spawn capacity-block purchase <offering-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--count		int32	1	Number of instances in the block
--dry-run		bool		Preview the price and terms without purchasing (no charge, no write API call)
--duration-hours		int32		Capacity Block duration in hours (required)
--instance-type		string		Instance type of the offering, e.g. p5.48xlarge (required)
--platform		string	Linux/UNIX	Instance platform (Linux/UNIX, Windows, ...)
--region		string		AWS region of the offering (required)
--tag		stringArray		Tag to apply to the reservation (key=value; repeatable)

spawn cleanup

Remove the shared AWS resources spore.host created (tagged spawn:managed), in dependency order. Running instances are NEVER removed — stop or terminate them first.

Preview what would be removed with --dry-run; otherwise cleanup prompts for confirmation (skip with --yes) and then deletes. By default it acts only on resources you created; --all widens to every principal in the account.

A log of everything removed is written to ~/.spawn/cleanup-<timestamp>.log.

spawn cleanup [flags]

Flags:

Flag	Short	Type	Default	Description
--all-regions		bool		Clean up every enabled region
--all		bool		Include resources created by other principals (default: only yours)
--dry-run		bool		Preview what would be removed without deleting anything
--region		string		AWS region (default: current region from AWS config)
--yes	-y	bool		Skip the confirmation prompt

spawn collect-results

Deprecated: use 'spawn sweep collect <sweep-id>' instead

Collect results from all instances in a parameter sweep.

This command downloads result files from S3 (uploaded by sweep instances), aggregates them into a single output file, and optionally identifies the best performing parameters based on a metric.

Result File Convention: Instances should upload results to: s3://spawn-results-<account>-<region>/sweeps/<sweep-id>/<index>/results.json

The result file should be a JSON object with metrics, e.g.: { "accuracy": 0.95, "loss": 0.12, "duration": 120.5, "params": {...} }

Examples:

Collect all results to JSON

spawn collect-results --sweep-id sweep-123 --output results.json

Collect to CSV format

spawn collect-results --sweep-id sweep-123 --output results.csv --format csv

Find top 5 runs by accuracy (descending)

spawn collect-results --sweep-id sweep-123 --metric accuracy --best 5

Custom S3 prefix (if instances uploaded to different location)

spawn collect-results --sweep-id sweep-123 --s3-prefix s3://my-bucket/results/

spawn collect-results [flags]

Flags:

Flag	Short	Type	Default	Description
--best		int		Show only top N results by metric (0 = all)
--format		string	json	Output format: json, csv, jsonl
--metric		string		Metric to use for ranking results (e.g., accuracy, loss)
--output	-o	string	results.json	Output file path
--regions	-r	stringSlice		Regions to collect from (comma-separated or repeated; default: all)
--s3-prefix		string		Custom S3 prefix for results (default: auto-detect)
--sweep-id		string		Sweep ID to collect results from (required)

spawn connect

Aliases: ssh

Connect to a spawn-managed instance via SSH.

Automatically finds your SSH key and connects. Falls back to AWS Session Manager if SSH is unavailable.

Examples:

Connect by instance ID

spawn connect i-1234567890abcdef0

Connect by name

spawn connect my-instance

spawn connect <instance-id> [-- <command>...] [flags]

Flags:

Flag	Short	Type	Default	Description
--key		string		SSH private key path
--no-start		bool		Do not automatically start a stopped/hibernated instance
--port		int	22	SSH port
--rdp-port		int	13389	Windows --rdp --via-ssm: local port for the SSM RDP tunnel
--rdp		bool		Windows: open a Remote Desktop (RDP) connection (decrypts the Administrator password)
--session-manager		bool		Use AWS Session Manager instead of SSH
--ssh		bool		Windows: SSH in (as Administrator, over OpenSSH) instead of opening a PowerShell-over-SSM session — same SSH path as Linux
--user		string		SSH username (default: ec2-user)
--via-ssm		bool		Windows --rdp: tunnel RDP over an SSM port-forwarding session instead of connecting to the public IP

spawn cost

Display detailed cost breakdown by region and instance type.

Shows:

• Resource costs (compute, storage, network)
• Cloud economics (effective cost/hr, utilization, savings)
• Time breakdown (running vs stopped hours)
• Budget status (if budget was set)
• Cost by region and instance type

Examples: spawn cost sweep-20260124-140530

spawn cost <sweep-id>

spawn create-ami

Deprecated: use 'spawn ami create' instead

Create an AMI from a running instance with automatic tagging.

The AMI will be tagged with spawn metadata for easy discovery and management.

Examples:

Create AMI from instance

spawn create-ami my-instance --name pytorch-2.2-cuda12

With custom tags

spawn create-ami i-abc123
--name my-stack-v1.0
--description "My custom software stack"
--tag stack=myapp
--tag version=1.0
--tag gpu=true

Wait for AMI to be available

spawn create-ami my-instance --name my-ami --wait

Allow reboot (default is no-reboot)

spawn create-ami my-instance --name my-ami --reboot

spawn create-ami <instance-id-or-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--description		string		Description for the AMI
--name		string		Name for the AMI (required)
--reboot		bool		Reboot instance before creating AMI (default: no-reboot)
--tag		stringArray		Tags in key=value format (can be specified multiple times)
--wait		bool		Wait for AMI to become available

spawn defaults

Manage default values for spawn launch flags.

Defaults are stored in ~/.spawn/config.yaml and applied whenever the corresponding flag is not explicitly provided on the command line.

Valid keys: slack-workspace Slack workspace ID for lifecycle notifications (e.g. T03NE3GTY) active-processes Process names to monitor for idle detection (e.g. rsession) active-ports TCP ports to monitor for active connections (e.g. 8787) idle-timeout Default idle timeout duration (e.g. 30m, 1h) hibernate-on-idle Hibernate instead of terminating on idle (true/false)

Examples: spawn defaults set slack-workspace T03NE3GTY spawn defaults set active-processes rsession spawn defaults set idle-timeout 1h spawn defaults list spawn defaults unset active-processes

spawn defaults

spawn defaults list

List all default launch values

spawn defaults list

spawn defaults set

Set a default launch value

spawn defaults set <key> <value>

spawn defaults unset

Remove a default launch value

spawn defaults unset <key>

spawn dns

Manage DNS records for spawn instances.

DNS names are automatically registered when launching with --dns flag. Format: <name>.<account-base36>.spore.host

Examples:

List DNS-enabled instances

spawn dns list

Register DNS name

spawn dns register i-1234567890abcdef0 my-server

Delete DNS record

spawn dns delete my-server

spawn dns

Flags:

Flag	Short	Type	Default	Description
--domain		string		DNS domain for record registration (default: spore.host)

spawn dns delete

spawn dns delete <instance-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn dns list

spawn dns list [flags]

Flags:

Flag	Short	Type	Default	Description
--all	-a	bool		Show all instances (including those without DNS)

spawn dns register

spawn dns register <instance-id> <dns-name>

spawn extend

Extend the TTL (time-to-live) for a spawn-managed instance.

Prevents automatic termination by extending the TTL duration.

Duration format: 1h, 2h30m, 24h, etc.

Examples:

Extend by 2 hours

spawn extend i-1234567890abcdef0 2h

Extend by name

spawn extend my-instance 8h

spawn extend <instance-id-or-name> <duration> [flags]

Flags:

Flag	Short	Type	Default	Description
--job-array-id		string		Extend TTL for all instances in job array by ID
--job-array-name		string		Extend TTL for all instances in job array by name

spawn fsx

Manage FSx Lustre filesystems: list, info, export, delete

spawn fsx

spawn fsx delete

Delete an FSx Lustre filesystem, optionally exporting to S3 first

spawn fsx delete <filesystem-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--export-first		bool		Export data to S3 before deleting
--yes	-y	bool		Skip confirmation prompt

spawn fsx list

List all spawn-managed FSx Lustre filesystems across all regions

spawn fsx list

spawn fsx show

Aliases: info

Show detailed information about an FSx Lustre filesystem

spawn fsx show <filesystem-id>

spawn hibernate

Hibernate an instance to disk (saves RAM state)

spawn hibernate [instance-id-or-name] [flags]

Flags:

Flag	Short	Type	Default	Description
--job-array-id		string		Hibernate all instances in job array by ID
--job-array-name		string		Hibernate all instances in job array by name

spawn image

Create custom AMIs that spawn can launch.

Currently supports importing a Windows 11 ISO into an AMI via AWS EC2 Image Builder's managed import-disk-image workflow (drivers, EC2Launch, SSM agent and Defender are pre-staged automatically). See infra/amis/windows/README.md.

spawn image

spawn image import

Convert a Windows 11 ISO into an AMI using EC2 Image Builder's managed import-disk-image workflow, then tag it so 'spawn launch --os windows' can use it.

The ISO must be a SUPPORTED, NON-evaluation Windows 11 Enterprise image (23H2 / 24H2 / 25H2 x64) obtained from the Microsoft 365 admin center. Evaluation, Media-Creation-Tool, and LTSC ISOs are rejected by the service. Bring your own Microsoft license (BYOL).

The command self-provisions the IAM roles and Image Builder infrastructure configuration it needs (idempotent); pass --infra-config-arn only to reuse an existing/custom one. See infra/amis/windows/README.md.

Examples:

Local ISO — staging bucket + infra auto-provisioned, nothing to pre-create:

spawn image import --iso ./Win11_25H2_Enterprise.iso
--name win11-25h2 --image-index 3

ISO already in S3 (uppercase .ISO key required by the service):

spawn image import --iso s3://my-bucket/Win11_25H2_Enterprise.ISO
--name win11-25h2

spawn image import [flags]

Flags:

Flag	Short	Type	Default	Description
--bucket		string		S3 bucket to stage a local ISO in (default: managed spawn-iso-import-<account>-<region>, auto-created)
--execution-role		string	AWSServiceRoleForImageBuilder	IAM execution role name or ARN
--image-index		int64	1	1-based edition index in a multi-edition ISO
--infra-config-arn		string		Image Builder infrastructure configuration ARN (optional; self-provisioned if omitted)
--instance-type		string	m6i.large	Build instance type (when self-provisioning infra)
--iso		string		Windows 11 ISO: local path or s3://bucket/key.ISO (required)
--keep-iso		bool		Keep the staged ISO (and managed bucket) after the AMI is built; by default they are deleted (only applies with --wait)
--name		string		Image Builder image resource name (required)
--no-secure-boot		bool		Disable Secure Boot on the output AMI
--no-warm		bool		Skip building the warm (fast-boot) AMI; produce only the raw imported base AMI
--region		string	us-east-1	AWS region for the import build
--s3-key		string		S3 object key for the uploaded ISO (default: derived from filename, .ISO)
--security-group-ids		stringSlice		Security groups for the build instance (comma-separated or repeated; when self-provisioning infra)
--subnet-id		string		Subnet for the build instance (when self-provisioning infra)
--version		string	1.0.0	Semantic version for the output image (major.minor.patch)
--wait-timeout		int	60	Max minutes to wait when --wait is set before detaching (the build keeps running)
--wait		bool		Wait for the AMI to finish building, then tag and clean up (warm mode implies this)
--warm-instance-type		string	m7i.xlarge	Instance type for the warm-build seed (non-burstable; Windows)
--warm-timeout		int	30	Safety-net minutes to wait for the warm seed's first boot (Administrator password) before giving up

spawn image status

Report the current state of an EC2 Image Builder image build started by 'spawn image import' (without --wait). Prints PENDING/BUILDING/.../AVAILABLE/FAILED and, once available, the output AMI id.

Example: spawn image status arn:aws:imagebuilder:us-east-1:123456789012:image/win11-25h2/1.0.0/1

spawn image status <image-build-version-arn> [flags]

Flags:

Flag	Short	Type	Default	Description
--region		string	us-east-1	AWS region of the image build

spawn image verify

Inspect a local Windows installation ISO and report which editions it contains and whether 'spawn image import' (EC2 Image Builder import-disk-image) will accept it — before you spend a real, paid build.

import-disk-image accepts only Windows 11 Enterprise (23H2/24H2/25H2, x64), non-Evaluation. This reads the ISO's install.wim metadata directly (no mount, no external tools) and prints each edition with its image index, flags the one to use, and gives a clear ACCEPTED/REJECTED verdict.

Examples: spawn image verify "/Volumes/External HD/Win11_Enterprise_25H2.iso" spawn image verify win11.iso -o json

spawn image verify <path-to.iso>

spawn instance-config

Aliases: config

Read or write runtime config on a running instance via SSH

spawn instance-config <instance-id> <action> [key] [value]

spawn launch

Aliases: , run, create

Launch an EC2 instance with smart defaults.

Three ways to use:

1. Interactive wizard (default if no input)
2. From truffle JSON via pipe
3. Direct with flags

Examples:

Interactive wizard

spawn launch

From truffle

truffle search m7i.large | spawn launch

Direct

spawn launch --instance-type m7i.large --region us-east-1

spawn launch <name> [flags]

Flags:

Flag	Short	Type	Default	Description
--active-ports		string		TCP ports to monitor for active connections, prevents idle termination (e.g. '8787' for RStudio, '8787,8888' for RStudio+Jupyter)
--active-processes		string		Process names to monitor, prevents idle termination while any are running (e.g. 'rsession' for RStudio, 'rsession,jupyter' for multiple)
--allow-cidr		string		CIDR allowed to reach the managed Windows security group (RDP 3389 + SSH 22); default 0.0.0.0/0
--ami		string		AMI ID (ami-...); omit or use 'auto' to auto-detect the latest AL2023
--attach-volume		stringArray		Attach an EBS volume from a snapshot, mounted at a path: snap-xxx:/mount/point[:ro]. Repeatable. Read-only is the common case for shared reference data.
--auto-placement-group		bool	true	Automatically create placement group for MPI job arrays (default: true)
--az		string		Availability zone
--batch-queue		string		Batch job queue file (JSON) for sequential execution
--budget		float64		Budget limit in dollars for parameter sweeps (0 = no limit)
--capacity-block		bool		The --reservation-id is a Capacity Block for ML (sets MarketType=capacity-block); mutually exclusive with --spot (#216)
--cartesian		bool		Generate cartesian product of parameter lists
--command		string		Command to run on all instances (executed after spored setup)
--completion-delay		string	30s	Grace period after completion signal
--completion-file		string	/tmp/SPAWN_COMPLETE	File to watch for completion signal
--compliance-strict		bool		Strict mode: fail on warnings (default: show warnings only)
--config		string		Launch config YAML file (supports plugins: list)
--cost-limit		float64		Terminate/stop when compute spend reaches this amount in USD (compute cost only; 0 = disabled)
--cost-tier		string		Prefer cost tier: low, standard, premium
--count		int	1	Number of instances to launch (job array)
--detach		bool		Run sweep orchestration in Lambda (auto-enabled for parameter sweeps)
--dns-api-endpoint		string		Custom DNS API endpoint (overrides default)
--dns-domain		string		Custom DNS domain (overrides default)
--dns		string		Override DNS name if different from --name (advanced)
--efa		bool		Enable Elastic Fabric Adapter for ultra-low latency MPI (requires supported instance types)
--efs-id		string		EFS filesystem ID to mount (fs-xxx)
--efs-mount-options		string		Custom EFS mount options (overrides profile)
--efs-mount-point		string	/efs	EFS mount point (default: /efs)
--efs-profile		string	general	EFS performance profile: general, max-io, max-throughput, burst
--estimate-only		bool		Show cost estimate and exit without launching
--fsx-create		bool		Create new FSx Lustre filesystem with S3 backing (requires --fsx-lifecycle)
--fsx-export-path		string		S3 path to export to (e.g., s3://bucket/prefix)
--fsx-id		string		Existing FSx Lustre filesystem ID to mount (fs-xxx)
--fsx-import-path		string		S3 path to import from (e.g., s3://bucket/prefix)
--fsx-lifecycle		string		FSx lifetime (REQUIRED with --fsx-create): 'ephemeral' (reaped when this instance terminates) or 'durable' (persists; requires --fsx-ttl)
--fsx-mount-point		string	/fsx	FSx mount point (default: /fsx)
--fsx-recall		string		Recall FSx filesystem by stack name (recreate from S3)
--fsx-s3-bucket		string		S3 bucket for FSx import/export (required with --fsx-create)
--fsx-skip-validate		bool		Skip FSx filesystem validation (for testing)
--fsx-storage-capacity		int32	1200	FSx storage capacity in GB (1200, 2400, or increments of 2400)
--fsx-throughput		int32	125	FSx PERSISTENT_2 throughput in MB/s/TiB (125, 250, 500, or 1000; default: 125)
--fsx-ttl		string		FSx time-to-live, required for --fsx-lifecycle=durable (e.g. 7d, 720h) — the filesystem is reaped this long after creation once no instance is using it
--hibernate		bool		Enable hibernation
--iam-allow-full-access		bool		Permit wildcard :FullAccess --iam-policy templates (s3:/dynamodb:/sqs: on all resources) on the instance role; off by default — prefer scoped ReadOnly/WriteOnly
--iam-managed-policies		stringSlice		AWS managed policy ARNs
--iam-policy-file		string		Custom IAM policy JSON file
--iam-policy		stringSlice		Service-level policies (e.g., s3:ReadOnly,dynamodb:WriteOnly). Wildcard *:FullAccess templates require --iam-allow-full-access
--iam-role-tags		stringSlice		Tags for IAM role (key=value format)
--iam-role		string		IAM role name (creates if doesn't exist)
--iam-trust-services		stringSlice	[ec2]	Services that can assume role
--idle-timeout		string		Auto-terminate if idle (defaults to 1h if neither --ttl nor --idle-timeout set)
--instance-names		string		Instance name template (e.g., 'worker-{index}', default: '{job-array-name}-{index}')
--instance-type		string		Instance type
--interactive		bool		Force interactive wizard
--job-array-name		string		Job array group name (required if --count > 1)
--key-name		string		SSH key pair name (EC2 KeyName)
--launch-delay		string	0s	Delay between instance launches (e.g., 5s)
--max-concurrent-per-region		int		Max instances running simultaneously per region (0 = unlimited)
--max-concurrent		int		Max instances running simultaneously (0 = unlimited)
--min-viable		int	1	Job array: minimum members that must launch for success (default 1; ignored for --mpi)
--mode		string	balanced	Distribution mode: balanced (fair share) or opportunistic (prioritize available regions)
--mpi-command		string		Command to run via mpirun (alternative to --command)
--mpi-processes-per-node		int		MPI processes per node (default: vCPU count)
--mpi		bool		Enable MPI cluster setup (requires --count > 1)
--name		string		Name your spore, required (sets Name tag, DNS, and hostname)
--nested-virtualization		bool		Enable nested virtualization (run KVM/Hyper-V inside the instance). Requires a C8i/M8i/R8i instance type.
--nist-800-171		bool		Enable NIST 800-171 Rev 3 compliance mode
--nist-800-53		string		Enable NIST 800-53 compliance (low, moderate, high)
--no-detach		bool		Disable auto-detach for parameter sweeps (requires --ttl or --idle-timeout)
--no-timeout		bool		Disable automatic timeout (NOT RECOMMENDED: creates zombie risk)
--notify-platform		string		Chat platform for lifecycle notifications: slack (default), teams, or discord
--on-complete		string		Action when workload signals completion: terminate, stop, hibernate. Use 'terminate' for batch/headless workloads — 'stop' leaves EBS (and any attached EIP) billing indefinitely, which is easy to forget in accounts without a hosted reaper
--on-idle		string		Action when the instance goes idle: stop (default) or hibernate. Mirrors --on-complete. NOTE: a stopped/hibernated instance keeps billing for its EBS volumes (and any attached Elastic IP) — for batch/headless work prefer --on-complete terminate so cost is fully bounded
--os		string		Target OS: windows or linux. Omit to auto-detect from the AMI. Use to force the OS for a custom AMI whose platform metadata is unset.
--output-id		string		Write sweep/instance ID to file for scripting
--param-file		string		Path to parameter sweep file (JSON/YAML/CSV)
--params		string		Inline JSON parameters for sweep
--placement-group		string		AWS Placement Group for MPI instances (auto-created if not specified)
--plugin		stringArray		Plugin to install at launch (ref[@version], repeatable)
--pre-stop-timeout		string		Max time to wait for --pre-stop command (default: 5m, spot: 90s)
--pre-stop		string		Shell command to run on the instance before any lifecycle-triggered stop/terminate (e.g., "aws s3 sync /results s3://bucket/")
--proximity-from		string		Prefer regions close to this region (e.g., us-east-1)
--queue-template		string		Queue template name (use 'spawn queue template list' to see options)
--quiet		bool		Minimal output
--region		string		AWS region
--regions-exclude		stringSlice		Exclude these regions (supports wildcards: us-, eu-)
--regions-geographic		stringSlice		Geographic constraints: us, eu, ap, north-america, europe, asia-pacific
--regions-include		stringSlice		Only use these regions (supports wildcards: us-, eu-)
--reservation-id		string		Capacity Reservation / Capacity Block ID to launch into (fs-/cr-...) — instance must be in the reservation's AZ (#216)
--security-group-ids		stringSlice		Security group IDs (comma-separated or repeated)
--session-timeout		string	30m	Auto-logout idle shells (0 to disable)
--skip-mpi-install		bool		Skip MPI installation (use with custom AMIs that have MPI pre-installed)
--skip-region-check		bool		Skip data locality region mismatch warnings
--slack-workspace		string		Slack workspace ID for lifecycle notifications (e.g. T03NE3GTY)
--spot-max-price		string		Max Spot price
--spot-webhook-url		string		On spot interruption, spored POSTs a fire-once, best-effort notice to this URL within the ~2-min window (off-node consumers; empty = disabled)
--spot		bool		Launch as Spot instance
--strata-formation		string		Strata formation to activate (e.g. r-research@2024.03)
--strata-profile		string		Path to a Strata profile YAML file
--strata-registry		string	s3://strata-registry	Strata registry S3 URL
--subnet-id		string		Subnet ID
--sweep-name		string		Human-readable sweep identifier (auto-generated if empty)
--tag		stringArray		Custom tag key=value on the instance and its created volumes (repeatable). The spawn: prefix is reserved.
--team		string		Team ID: tag instance with spawn:team-id for team-shared access
--template-var		stringToString		Template variables (key=value)
--terminate-on-error		bool		If post-launch verification fails (e.g. spored didn't come up), terminate the instance instead of leaving it running
--ttl		string		Auto-terminate after duration (e.g., 8h, defaults to 1h idle if not set)
--use-reservation		bool		Use capacity reservation
--user-data-file		string		User data file
--user-data		string		User data (@file or inline)
--volume-size		int32		Root EBS volume size in GiB (0 = use AMI default)
--vpc		string		VPC ID
--wait-for-running		bool	true	Wait until running
--wait-for-ssh		bool	true	Wait until SSH is ready
--wait-timeout		string		Timeout for --wait (e.g., 2h, 30m, 0=no timeout)
--wait		bool		Wait for sweep/launch to complete (requires --detach)
--webhook-correlation		string		Opaque blob echoed verbatim in the spot-webhook payload so a consumer can correlate the event to its own record (never parsed by spawn)
--webhook-timeout		string		Hard cap on the spot-webhook POST so it can't eat the reclamation window (default: 2s)
--yes	-y	bool		Auto-approve cost estimate (skip confirmation)

spawn list-amis

Deprecated: use 'spawn ami list' instead

List AMIs created and managed by spawn.

Filters AMIs by spawn tags to show only those created by spawn. You can filter by stack, version, architecture, and other attributes.

Examples:

List all spawn AMIs

spawn list-amis

Filter by stack

spawn list-amis --stack pytorch

Filter by stack and version

spawn list-amis --stack pytorch --version 2.2

Filter by architecture

spawn list-amis --arch arm64

Show only GPU AMIs

spawn list-amis --gpu true

Show deprecated AMIs

spawn list-amis --deprecated

JSON output

spawn list-amis --json

spawn list-amis [flags]

Flags:

Flag	Short	Type	Default	Description
--arch		string		Filter by architecture (x86_64 or arm64)
--deprecated		bool		Show deprecated AMIs (default: hide deprecated)
--gpu		string		Filter by GPU support (true or false)
--region		string		AWS region (default: current region from AWS config)
--stack		string		Filter by stack (spawn:stack tag)
--version		string		Filter by version (spawn:version tag)

spawn list-sweeps

Deprecated: use 'spawn sweep list' instead

List parameter sweeps from DynamoDB orchestration table.

Shows recent sweeps with their status, progress, and creation time.

Examples:

List recent sweeps

spawn list-sweeps

Filter by status

spawn list-sweeps --status RUNNING

Show last 5 sweeps

spawn list-sweeps --last 5

Show sweeps since a date

spawn list-sweeps --since 2026-01-15

JSON output

spawn list-sweeps --json

spawn list-sweeps [flags]

Flags:

Flag	Short	Type	Default	Description
--last		int	20	Show last N sweeps
--region		string		Filter by region
--since		string		Show sweeps created after date (YYYY-MM-DD)
--status		string		Filter by status (RUNNING, COMPLETED, FAILED, CANCELLED)

spawn list

Aliases: ls

List spawn-managed EC2 instances across regions.

Shows all instances with the spawn:managed tag.

Examples:

List all instances

spawn list

Filter by region

spawn list --region us-east-1

Filter by state

spawn list --state running

JSON output

spawn list --output json

spawn list [flags]

Flags:

Flag	Short	Type	Default	Description
--az		string		Filter by availability zone
--instance-family		string		Filter by instance family (e.g., m7i, t3)
--instance-type		string		Filter by exact instance type (e.g., t3.micro)
--job-array-id		string		Filter by job array ID
--job-array-name		string		Filter by job array name
--region		string		Filter by AWS region (default: all regions)
--regions	-r	stringSlice		Filter by regions (comma-separated, e.g. us-east-1,us-west-2)
--state		string		Filter by instance state (running, stopped, etc.)
--sweep-id		string		Filter by parameter sweep ID
--sweep-name		string		Filter by parameter sweep name
--tag		stringArray		Filter by tag (key=value format, can be specified multiple times)

spawn notify

Aliases: bot

Register and manage Slack/Teams/SMS notifications for instances.

Lets authorized users receive lifecycle events (launch, idle stop, TTL warn, termination) and control instances via chat slash commands without CLI access.

Examples: spawn notify register --platform slack --user professor@example.com
--instance i-0abc123 --nickname rstudio --allow start,stop,status spawn notify deregister --platform slack --user professor@example.com --nickname rstudio spawn notify list --platform slack --workspace T03NE3GTY

spawn notify

spawn notify deregister

Remove a chat bot registration

spawn notify deregister [flags]

Flags:

Flag	Short	Type	Default	Description
--nickname		string		Nickname to deregister
--platform		string		Chat platform: slack, teams, or discord
--table		string		Override DynamoDB registry table name
--user-id		string		Platform user ID
--workspace-id		string		Platform workspace ID

spawn notify disable

Suspend bot access without removing the registration. Use during sensitive computation runs or maintenance. Re-enable with 'spawn notify enable'.

spawn notify disable [flags]

Flags:

Flag	Short	Type	Default	Description
--nickname		string		Nickname of the registration to enable/disable
--platform		string		Chat platform: slack, teams, or discord
--table		string		Override DynamoDB registry table name
--user-id		string		Platform user ID
--workspace-id		string		Platform workspace ID

spawn notify enable

Grant bot access to a registered instance. Registrations are created disabled by default — this command must be run before a chat user can control the instance via slash commands.

spawn notify enable [flags]

Flags:

Flag	Short	Type	Default	Description
--nickname		string		Nickname of the registration to enable/disable
--platform		string		Chat platform: slack, teams, or discord
--table		string		Override DynamoDB registry table name
--user-id		string		Platform user ID
--workspace-id		string		Platform workspace ID

spawn notify list

List chat bot registrations for a workspace

spawn notify list [flags]

Flags:

Flag	Short	Type	Default	Description
--platform		string		Chat platform: slack, teams, or discord
--table		string		Override DynamoDB registry table name
--workspace-id		string		Platform workspace ID

spawn notify register

Register an EC2 instance so a chat user can control it via slash commands.

Supports specifying the user by email (--user) which resolves to a platform user ID, or directly by platform ID (--user-id + --workspace-id).

The --nickname is the friendly name used in slash commands, e.g.: /prism stop rstudio /prism status jupyter

Both the instance ID and instance name (DNS name or spawn:name tag) are accepted as the target in slash commands once registered.

spawn notify register [flags]

Flags:

Flag	Short	Type	Default	Description
--allow		stringSlice		Allowed actions (default: start,stop,status,hibernate,url)
--connect-code		string		One-time code from /spore connect (alternative to --user-id)
--instance		string		Instance ID (i-...) or name
--nickname		string		Friendly name for slash commands (default: 'default')
--platform		string		Chat platform: slack, teams, or discord
--role-arn		string		Cross-account IAM role ARN for this instance's account (created automatically if omitted)
--table		string		Override DynamoDB registry table name
--tag-prefix		string		Tag prefix: spawn or prism (default: auto-detected)
--user-id		string		Platform-native user ID (e.g. Slack U04KZABCD)
--user		string		User email address (resolved to platform user ID)
--workspace-id		string		Platform workspace ID (e.g. Slack T03NE3GTY)

spawn notify workspace

Manage chat-platform workspace registrations

spawn notify workspace

spawn notify workspace add

Store the Slack bot token and signing secret for a workspace so the spore-bot Lambda can verify incoming slash command requests.

Run this once after installing the Slack app in a workspace:

spawn notify workspace-add
--platform slack
--workspace-id T03NE3GTY
--workspace-name "My Workspace"
--bot-token xoxb-...
--signing-secret abc123...

spawn notify workspace add [flags]

Flags:

Flag	Short	Type	Default	Description
--allowed-channels		stringSlice		Restrict commands to specific channel IDs (e.g. C12345,C67890). Empty = all channels.
--bot-token		string		Bot token (Slack xoxb-..., or Discord bot token)
--connect-ttl		int		Max /spore connect code lifetime in hours (0 = use platform default, typically 24h). Can only lower the platform default.
--platform		string		Chat platform: slack, teams, or discord
--public-key		string		Discord application public key (Ed25519, hex; required for discord)
--signing-secret		string		Slack/Teams signing secret (required for slack/teams)
--table		string		Override DynamoDB workspaces table name
--webhook-url		string		Channel webhook URL for notifications (Discord channel webhook, or manual Slack incoming webhook)
--workspace-id		string		Platform workspace ID
--workspace-name		string		Human-friendly workspace name

spawn notify workspace destroy

Permanently delete all instance registrations across all users in a workspace, and remove the workspace's bot token and signing secret.

Preview with --dry-run; otherwise it prompts for confirmation (skip with --yes) and then executes the full teardown.

Note: The SpawnBotCrossAccount IAM role in customer accounts is not deleted automatically. Remove it separately with: aws cloudformation delete-stack --stack-name spawn-bot-cross-account

spawn notify workspace destroy [flags]

Flags:

Flag	Short	Type	Default	Description
--dry-run		bool		Preview what would be removed without deleting anything
--platform		string		Chat platform: slack, teams, or discord
--registry-table		string		Override DynamoDB registry table name
--workspace-id		string		Platform workspace ID (required)
--workspaces-table		string		Override DynamoDB workspaces table name
--yes	-y	bool		Skip the confirmation prompt

spawn notify workspace list

List registered workspaces

spawn notify workspace list [flags]

Flags:

Flag	Short	Type	Default	Description
--platform		string		Chat platform: slack, teams, or discord
--table		string		Override DynamoDB workspaces table name

spawn notify workspace remove

Remove a workspace registration

spawn notify workspace remove [flags]

Flags:

Flag	Short	Type	Default	Description
--platform		string		Chat platform: slack, teams, or discord
--table		string		Override DynamoDB workspaces table name
--workspace-id		string		Platform workspace ID
--yes	-y	bool		Skip the confirmation prompt

spawn orphans

Report spawn-managed resources that appear orphaned — present but with no running instance using them:

• EBS volumes in the 'available' state
• security groups not attached to any instance
• the shared infrastructure (key pair, IAM role) when no instances remain
• Elastic IPs that are unassociated, or attached to a stopped instance (an EIP keeps billing even while the instance is stopped)

This is a read-only report. Use 'spawn cleanup' to remove anything — except Elastic IPs: spawn never allocates them, so it never releases them. Any EIP listed is yours to release with 'aws ec2 release-address'.

spawn orphans [flags]

Flags:

Flag	Short	Type	Default	Description
--all-regions		bool		Search every enabled region
--all		bool		Include resources created by other principals (default: only yours)
--region		string		AWS region (default: current region from AWS config)

spawn pipeline

Manage multi-stage pipelines with DAG dependencies.

Pipelines allow you to orchestrate complex workflows where each stage runs on separate instances with different instance types. Stages can depend on each other, forming a directed acyclic graph (DAG).

Data can be passed between stages via:

• S3 (batch mode): Stage outputs uploaded to S3, downloaded by next stage
• Network streaming (real-time): Direct TCP/gRPC connections between stages

spawn pipeline

spawn pipeline cancel

Cancel a running pipeline and terminate all instances.

Sets the cancellation flag in DynamoDB. The orchestrator Lambda will terminate all running instances and mark the pipeline as CANCELLED.

spawn pipeline cancel <pipeline-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn pipeline collect

Download all results from a completed pipeline.

Downloads outputs from all stages to a local directory.

spawn pipeline collect <pipeline-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--output-dir		string	./results	Output directory for downloaded files
--stage		string		Download results from specific stage only

spawn pipeline graph

Display the pipeline dependency graph as ASCII art.

Shows:

• Stage names and instance types
• Dependencies between stages
• Fan-out and fan-in patterns
• Data passing modes (S3 or streaming)

spawn pipeline graph <file> [flags]

Flags:

Flag	Short	Type	Default	Description
--simple		bool		Show simplified graph
--stats		bool		Show graph statistics

spawn pipeline launch

Launch a multi-stage pipeline.

The pipeline definition will be uploaded to S3 and a Lambda orchestrator will be invoked to manage the pipeline execution.

spawn pipeline launch <file> [flags]

Flags:

Flag	Short	Type	Default	Description
--region		string		AWS region (default: from AWS config)
--wait		bool		Wait for pipeline to complete

spawn pipeline list

List all pipelines for the current user.

Shows pipeline ID, name, status, and cost.

spawn pipeline list [flags]

Flags:

Flag	Short	Type	Default	Description
--status		string		Filter by status (INITIALIZING, RUNNING, COMPLETED, FAILED, CANCELLED)

spawn pipeline status

Show the current status of a running or completed pipeline.

Displays:

• Overall pipeline status
• Per-stage progress
• Instance information
• Cost tracking

spawn pipeline status <pipeline-id>

spawn pipeline validate

Validate a pipeline definition file.

Checks:

• JSON syntax and structure
• Required fields present
• Stage dependencies valid (no circular dependencies)
• Instance types, regions, and other configuration valid

spawn pipeline validate <file>

spawn plugin

Install, inspect, and remove service plugins on running instances.

Plugins are composable service units (Jupyter, Globus, Tailscale, etc.) defined by YAML specs with install/start/stop/health lifecycles.

Examples: spawn plugin list --instance i-0abc123 spawn plugin install globus-personal-endpoint --instance i-0abc123
--config endpoint_name=my-endpoint spawn plugin status globus-personal-endpoint --instance i-0abc123 spawn plugin remove globus-personal-endpoint --instance i-0abc123

spawn plugin

spawn plugin gen-index

Generate index.json for a plugin registry from its plugins/ directory.

This is the generator side of plugin discovery, in the same arrangement as 'spawn plugin manifest': it lives here, and the registry's CI invokes it, so the index is always derived by the same parser that installs plugins and can never describe a spec differently from the spec itself.

--generated-at takes an RFC3339 timestamp (CI passes the commit time) so regenerating an unchanged registry produces a byte-identical file. It defaults to now, which makes every run differ — fine locally, churn in CI.

Examples: spawn plugin gen-index ./plugins -o index.json spawn plugin gen-index ./plugins --generated-at 2026-07-29T00:00:00Z

spawn plugin gen-index <plugins-dir> [flags]

Flags:

Flag	Short	Type	Default	Description
--generated-at		string		RFC3339 generation timestamp (default: now)
--output	-o	string		Write index to this file instead of stdout
--source		string	spore-host/spore-plugins	owner/repo the index describes

spawn plugin info

Show what the registry knows about a plugin: version, description, config parameters, and declared capability surface.

This reads the registry index and contacts no instance. It describes the plugin as PUBLISHED — for the full spec of what installing would run, including every step, use 'spawn plugin inspect <ref>'.

Examples: spawn plugin info tailscale spawn plugin info jupyterlab --output json

spawn plugin info <name> [flags]

Flags:

Flag	Short	Type	Default	Description
--refresh		bool		Bypass the local cache and refetch the registry index

spawn plugin inspect

Resolve a plugin reference and render its plan — resolved source and version, local (controller) vs remote (instance) steps, requested controller environment, root vs login-user execution, downloads, health checks, cleanup, and its declared permissions block — WITHOUT executing anything or contacting an instance.

Installing a plugin runs its author's code on your machine and, on the instance, as root. Inspect it first, especially for third-party (github:) plugins.

Plugin ref formats are the same as 'spawn plugin install': name official registry (spore-host/spore-plugins) name@v1.2.0 pinned to git tag github:user/repo/name custom GitHub repository ./path/to/plugin.yaml local file

spawn plugin inspect <plugin-ref> [flags]

Flags:

Flag	Short	Type	Default	Description
--insecure		bool		Skip signature/checksum verification for official plugin releases (unsafe)

spawn plugin install

Install a plugin on a running spore instance.

Runs the plugin's full lifecycle: local provision steps on this controller (e.g. creating a mutagen sync or a Globus endpoint), then the remote install/configure/start steps on the instance via spored. Values captured and pushed by local steps are delivered before the remote configure phase runs.

Requires SSH access to the instance (the same key used by 'spawn plugin status').

Plugin ref formats: name official registry (spore-host/spore-plugins) name@v1.2.0 pinned to git tag github:user/repo/name custom GitHub repository ./path/to/plugin.yaml local file

spawn plugin install <plugin-ref> [flags]

Flags:

Flag	Short	Type	Default	Description
--config		stringArray		Config as key=value (repeatable)
--dry-run		bool		Preview the plan without installing (contacts no instance)
--insecure		bool		Skip signature/checksum verification for official plugin releases (unsafe)
--instance	-i	string		Instance ID or hostname (required)
--key		string		Path to SSH private key
--user		string		SSH username for the instance (default: ec2-user)

spawn plugin list

List plugins installed on an instance

spawn plugin list [flags]

Flags:

Flag	Short	Type	Default	Description
--instance	-i	string		Instance ID or hostname (required)
--key		string		Path to SSH private key
--user		string		SSH username for the instance (default: ec2-user)

spawn plugin manifest

Generate the checksum manifest (manifest.json) for a plugin directory. The manifest records the sha256 of the plugin's plugin.yaml so that spawn can verify a fetched official plugin matches the released bytes. This is the generator side of the registry supply-chain story: the registry's release workflow runs it and publishes the output as a GitHub Release asset; spawn verifies against it at install time. Contacts nothing.

Examples: spawn plugin manifest ./plugins/tailscale spawn plugin manifest ./plugins/tailscale -o manifest.json

spawn plugin manifest <plugin-dir> [flags]

Flags:

Flag	Short	Type	Default	Description
--output	-o	string		Write manifest to this file instead of stdout

spawn plugin remove

Remove a plugin from an instance

spawn plugin remove <name> [flags]

Flags:

Flag	Short	Type	Default	Description
--instance	-i	string		Instance ID or hostname (required)
--key		string		Path to SSH private key
--user		string		SSH username for the instance (default: ec2-user)
--yes	-y	bool		Skip the confirmation prompt

spawn plugin search

Search the official plugin registry (spore-host/spore-plugins) for plugins available to install. With no query, lists everything.

Reads a generated index published by the registry, cached locally so this works offline; the age of what you're seeing is always shown. This lists what EXISTS — use 'spawn plugin list --instance <id>' for what is installed on an instance.

Examples: spawn plugin search spawn plugin search jupyter spawn plugin search --refresh

spawn plugin search [query] [flags]

Flags:

Flag	Short	Type	Default	Description
--refresh		bool		Bypass the local cache and refetch the registry index

spawn plugin status

Show status of a plugin on an instance

spawn plugin status <name> [flags]

Flags:

Flag	Short	Type	Default	Description
--instance	-i	string		Instance ID or hostname (required)
--key		string		Path to SSH private key
--user		string		SSH username for the instance (default: ec2-user)

spawn plugin validate

Statically validate one or more plugin.yaml files without contacting any instance. Checks schema, semver, known step/condition/config types, that the containing directory matches the plugin name, and that every {{ config.X }} template reference points at a declared config parameter.

With --strict, also enforce that the declared permissions: block is consistent with the plugin's steps (e.g. instance.root=false must have no remote step that runs as root). --strict requires a permissions: block. The official registry's CI runs --strict so a published plugin's declared capability surface is enforced.

Examples: spawn plugin validate ./plugins/tailscale/plugin.yaml spawn plugin validate --strict ./plugins/*/plugin.yaml

spawn plugin validate <path>... [flags]

Flags:

Flag	Short	Type	Default	Description
--strict		bool		Also enforce permissions/step consistency (requires a permissions: block)

spawn queue

Commands for managing and monitoring batch job queues.

Batch queues execute jobs sequentially on a single instance, with dependency management and automatic result collection.

Examples:

Check queue status on instance

spawn queue status i-1234567890abcdef0

Download queue results

spawn queue results queue-20260122-140530 --output ./results/

spawn queue

spawn queue results

Download all job results from S3 for a completed or running queue.

Results include job outputs, logs, and the final queue state.

Examples:

Download to current directory

spawn queue results queue-20260122-140530

Download to specific directory

spawn queue results queue-20260122-140530 --output ./my-results/

spawn queue results <queue-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--output-dir		string	.	Output directory for results

spawn queue status

Show the execution status of a batch queue running on an instance.

Connects to the instance via SSH and reads the queue state file.

Examples: spawn queue status i-1234567890abcdef0

spawn queue status <instance-id>

spawn queue template

Manage pre-built queue configuration templates.

Templates provide ready-to-use queue configurations for common workflows with variable substitution for customization.

Available commands: list - List available templates show - Show template details generate - Generate queue config from template

spawn queue template

spawn queue template generate

Generate a queue configuration file from a template with variable substitution.

Variables can be provided via --var flags or use template defaults.

Examples:

Generate with defaults, output to file

spawn queue template generate ml-pipeline --output pipeline.json

Provide required variables

spawn queue template generate ml-pipeline
--var INPUT=/data/train.csv
--var S3_BUCKET=my-results
--output pipeline.json

Output to stdout (for piping)

spawn queue template generate simple-sequential
--var S3_BUCKET=results

spawn queue template generate <template-name> [flags]

Flags:

Flag	Short	Type	Default	Description
--output-file		string		Output file (default: stdout)
--var		stringToString		Template variables (key=value)

spawn queue template init

Launch an interactive wizard to create a custom queue configuration.

Guides you through creating a queue by asking questions about:

• Workflow type and name
• Number of jobs and commands
• Job dependencies
• Timeouts and retry policies
• Result collection
• S3 bucket configuration

Examples: spawn queue template init spawn queue template init --output my-queue.json

spawn queue template init [flags]

Flags:

Flag	Short	Type	Default	Description
--output-file		string		Output file (default: queue.json)

spawn queue template list

List all available queue configuration templates.

Shows template names, descriptions, and required/optional variables.

Examples: spawn queue template list

spawn queue template list

spawn queue template show

Show detailed information about a queue template.

Displays template description, jobs, and all variables with their defaults.

Examples: spawn queue template show ml-pipeline spawn queue template show etl

spawn queue template show <template-name>

spawn resources

List every AWS resource spore.host created in an account/region, found by the spawn:managed=true tag via the Resource Groups Tagging API.

By default it lists resources created by you (your IAM principal). Use --all to include resources created by other principals in the account.

spawn resources [flags]

Flags:

Flag	Short	Type	Default	Description
--all-regions		bool		Search every enabled region
--all		bool		Include resources created by other principals (default: only yours)
--region		string		AWS region (default: current region from AWS config)

spawn resume

Deprecated: use 'spawn sweep resume <sweep-id>' instead

Resume an interrupted parameter sweep from checkpoint.

Reads the sweep state from ~/.spawn/sweeps/<sweep-id>.json, queries EC2 for current instance states, and continues launching pending parameter sets with rolling queue orchestration.

Examples:

Resume sweep with original settings

spawn resume --sweep-id hyperparam-20260115-abc123

Resume with different max-concurrent

spawn resume --sweep-id <id> --max-concurrent 5

Resume in detached mode (Lambda)

spawn resume --sweep-id <id> --detach

spawn resume [flags]

Flags:

Flag	Short	Type	Default	Description
--detach		bool		Run sweep orchestration in Lambda
--max-concurrent		int		Override max concurrent instances (0 = use original)
--sweep-id		string		Sweep ID to resume (required)

spawn schedule

Create, list, and manage scheduled executions of parameter sweeps.

Schedules run parameter sweeps at specified times via EventBridge Scheduler. No CLI running required - sweeps launch automatically at scheduled times.

Examples:

One-time execution

spawn schedule create params.yaml --at "2026-01-23T02:00:00" --timezone "America/New_York"

Recurring daily at 2 AM

spawn schedule create params.yaml --cron "0 2 * * *" --name "nightly-training"

Recurring with execution limit

spawn schedule create params.yaml --cron "0 */4 * * *" --max-executions 100

List all schedules

spawn schedule list

Cancel a schedule

spawn schedule cancel <schedule-id>

spawn schedule

spawn schedule cancel

Cancel a scheduled execution. This will delete the EventBridge schedule and update the DynamoDB record. No further executions will occur.

Examples: spawn schedule cancel sched-20260122-140530

spawn schedule cancel <schedule-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn schedule create

Create a new scheduled execution of a parameter sweep.

Either --at (one-time) or --cron (recurring) is required.

Time formats: --at: ISO 8601 format (2026-01-23T02:00:00) --cron: Standard cron expression (minute hour day month weekday)

Examples:

One-time at specific time

spawn schedule create params.yaml --at "2026-01-23T14:30:00"

Every day at 2 AM Eastern

spawn schedule create params.yaml --cron "0 2 * * *" --timezone "America/New_York"

Every 6 hours for 30 days

spawn schedule create params.yaml --cron "0 */6 * * *" --max-executions 120

Weekdays only at 9 AM

spawn schedule create params.yaml --cron "0 9 * * 1-5"

spawn schedule create <params-file> [flags]

Flags:

Flag	Short	Type	Default	Description
--at		string		One-time execution time (ISO 8601 format)
--cron		string		Cron expression for recurring execution
--end-after		string		Stop executing after this time (ISO 8601 format)
--max-executions		int		Maximum number of executions (0 = unlimited)
--name		string		Friendly name for this schedule
--region		string	us-east-1	AWS region for sweep execution
--timezone		string	UTC	IANA timezone (e.g., America/New_York)

spawn schedule list

List all scheduled executions for the current user.

Examples:

List all schedules

spawn schedule list

List only active schedules

spawn schedule list --status active

spawn schedule list [flags]

Flags:

Flag	Short	Type	Default	Description
--status		string		Filter by status (active|paused|cancelled)

spawn schedule pause

Pause a scheduled execution temporarily. The schedule remains but executions are disabled. Use 'resume' to re-enable.

Examples: spawn schedule pause sched-20260122-140530

spawn schedule pause <schedule-id>

spawn schedule resume

Resume a paused scheduled execution. Executions will continue according to the schedule.

Examples: spawn schedule resume sched-20260122-140530

spawn schedule resume <schedule-id>

spawn schedule show

Aliases: describe

Show detailed information about a scheduled execution including configuration, execution history, and next run time.

Examples: spawn schedule show sched-20260122-140530

spawn schedule show <schedule-id>

spawn slurm

Parse and convert Slurm batch scripts to spawn parameter files.

This enables HPC users to migrate existing Slurm workflows to the cloud with minimal changes. Supports array jobs, MPI jobs, and GPU jobs.

Examples:

Convert Slurm script to spawn parameters

spawn slurm convert job.sbatch --output params.yaml

Estimate cost before running

spawn slurm estimate job.sbatch

Convert and submit in one step

spawn slurm submit job.sbatch --spot

spawn slurm

spawn slurm convert

Parse a Slurm batch script and convert it to spawn parameter format.

The generated parameter file can be reviewed and edited before launching.

Supported Slurm directives: --array=N-M → Parameter sweep with M-N+1 tasks --time=HH:MM:SS → TTL for each instance --mem=XGB → Memory requirement for instance selection --cpus-per-task=N → CPU requirement for instance selection --gres=gpu:N → GPU requirement and instance selection --nodes=N → Multi-node MPI job (requires --mpi flag) --job-name=NAME → Instance name prefix

Custom #SPAWN directives (optional): #SPAWN --instance-type=TYPE → Override instance type selection #SPAWN --region=REGION → Override region #SPAWN --spot=true → Enable spot instances #SPAWN --ami=AMI_ID → Override AMI

Example: spawn slurm convert train.sbatch --output params.yaml spawn launch --params params.yaml

spawn slurm convert <script.sbatch> [flags]

Flags:

Flag	Short	Type	Default	Description
--output-file		string		Output parameter file (default: stdout)

spawn slurm estimate

Parse a Slurm batch script and estimate the cloud cost.

Provides a cost comparison between institutional cluster (free but queued) and cloud (paid but immediate).

Example: spawn slurm estimate train.sbatch

spawn slurm estimate <script.sbatch> [flags]

Flags:

Flag	Short	Type	Default	Description
--region		string		Region to price the estimate in (default: #SPAWN --region, else your AWS config region)

spawn slurm submit

Parse a Slurm batch script, convert to spawn parameters, and launch immediately.

This is a convenience command that combines 'convert' and 'launch' in one step. For complex jobs, consider using 'convert' first to review the generated parameters.

Example: spawn slurm submit train.sbatch --spot --yes

spawn slurm submit <script.sbatch> [flags]

Flags:

Flag	Short	Type	Default	Description
--region		string		Region to price the estimate in (default: #SPAWN --region, else your AWS config region)
--yes	-y	bool		Skip confirmation prompt

spawn snapshot

Create EBS snapshots from raw disk images without launching an EC2 instance, so large reference data (a Kraken2 DB, BLAST index, ML weights) can be attached to spores via 'spawn launch --attach-volume' instead of being baked into a custom AMI.

spawn snapshot

spawn snapshot create

Populate a new EBS snapshot directly using the EBS direct APIs — no EC2 instance and no attached volume. The result is a snapshot you can attach with 'spawn launch --attach-volume snap-xxx:/mount'.

--from accepts any of:

• a directory → its contents are packed into an ext4 filesystem image
• a .tar/.tar.gz/.tgz archive → unpacked into an ext4 filesystem image
• a raw disk image → streamed verbatim (its bytes ARE the block device)

Directories and tarballs are converted to ext4 in-process (pure Go — no mkfs and no builder instance), so this works the same from macOS, Linux, or Windows. The ext4 filesystem is sized to the data and capped at --size.

For a directory or tarball source, the ext4 image is built in a local temp file first (~the uncompressed data size — e.g. ~16 GB for a 16 GB DB) and removed when done; ensure that much free space, or use --temp-dir to point at a roomier disk. A raw image needs no scratch (it streams source → snapshot directly). Memory stays low regardless of size (the upload streams blocks concurrently, never buffering the whole image).

The upload sends the uncompressed image to AWS over your connection; for a large DB over a slow uplink, run this from AWS CloudShell or a small in-region EC2 instance so the upload is AWS-internal.

Examples:

From a directory:

spawn snapshot create --from ./kraken2-db/ --size 20 --name kraken2-k2pluspf

From a tarball (local or in S3):

spawn snapshot create --from ./k2_pluspf.tar.gz --size 20 --name kraken2 spawn snapshot create --from s3://genome-idx/k2_pluspf.tar.gz --size 20
--name kraken2 --region us-east-1

From a raw filesystem image you already built:

spawn snapshot create --from ./kraken2.raw --size 20 --name kraken2

spawn snapshot create [flags]

Flags:

Flag	Short	Type	Default	Description
--description		string		Snapshot description
--encrypted		bool		Create an encrypted snapshot
--from		string		Source: a directory, a .tar/.tar.gz/.tgz, or a raw disk image — local path or s3://bucket/key (required)
--kms-key		string		Customer-managed KMS key ARN for encryption (implies --encrypted)
--name		string		Name tag for the snapshot (also sets spawn:snapshot-name)
--region		string		AWS region (default: the configured region)
--size		int64		Volume size in GiB the snapshot is built for; the image must fit (required)
--tag		stringArray		Custom tag key=value to set on the snapshot (repeatable). Merged with the spawn:* baseline; cannot override a spawn: tag.
--temp-dir		string		Directory for the temporary ext4 image built from a dir/tarball source (default: system temp). Point at a roomy disk for large data.

spawn snapshot mount

Convenience for the head-node side of the reference-data-volume recipe: create an EBS volume from a snapshot, attach it to the instance this command runs on, and mount it (read-only by default) at <mount-point>.

This only works when run ON an EC2 instance (it identifies itself via IMDS). It's the one-command equivalent of: aws ec2 create-volume --snapshot-id … && aws ec2 attach-volume … && sudo mount -o ro …. Use it on a spawn head node (or any EC2 box running 'nextflow run') so an nf-core pipeline's head-side db_path validation finds the DB. Tasks don't need this — 'spawn launch --attach-volume' mounts the volume on each task automatically.

Example: sudo spawn snapshot mount snap-0abc123 /opt/databases/kraken2

spawn snapshot mount <snapshot-id> <mount-point> [flags]

Flags:

Flag	Short	Type	Default	Description
--rw		bool		Mount read-write (default: read-only — the right choice for shared reference data)

spawn stage

Stage data to regional S3 buckets for efficient multi-region parameter sweeps.

Data staging enables cost-optimized data movement by:

• Replicating data once to regional buckets
• Allowing instances to download from local region (free)
• Avoiding repeated cross-region transfers ($0.09/GB)

Cost savings example: 100GB dataset, 2 regions, 10 instances each:

• Without staging: $90.00 (cross-region transfers)
• With staging: $6.60 (one-time replication)
• Savings: $85.70 (93% reduction)

Commands: spawn stage upload <path> Stage data to regional buckets spawn stage list List staged data spawn stage estimate Estimate staging cost savings spawn stage delete <id> Delete staged data

spawn stage

spawn stage delete

Delete staged data from all regions and remove metadata.

spawn stage delete <staging-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn stage estimate

Estimate the cost difference between: A) Single-region storage with cross-region transfers B) Regional replication with local transfers

This helps determine if staging is cost-effective for your workload.

Example: spawn stage estimate
--data-size-gb 100
--instances 10
--regions us-east-1,us-west-2

spawn stage estimate [flags]

Flags:

Flag	Short	Type	Default	Description
--data-size-gb		int	100	Dataset size in GB
--instances		int	10	Number of instances per region
--regions	-r	stringSlice	[us-east-1,us-west-2]	Regions to estimate for (comma-separated or repeated)

spawn stage list

List all data currently staged in regional buckets.

spawn stage list

spawn stage upload

Upload a file or directory to spawn data staging buckets across regions.

The data will be:

1. Uploaded to the primary region
2. Replicated to additional regions
3. Tracked in DynamoDB for lifecycle management
4. Automatically deleted after 7 days

Example: spawn stage upload ./reference-genome.fasta
--regions us-east-1,us-west-2
--dest /mnt/data/reference.fasta
--sweep-id sweep-abc123

spawn stage upload <local-path> [flags]

Flags:

Flag	Short	Type	Default	Description
--dest		string		Destination path on instances (default: /mnt/data/<filename>)
--regions	-r	stringSlice	[us-east-1,us-west-2]	Regions to replicate to (comma-separated or repeated)
--sweep-id		string		Associate with sweep ID for tracking

spawn start

Start a stopped or hibernated instance

spawn start [instance-id-or-name] [flags]

Flags:

Flag	Short	Type	Default	Description
--job-array-id		string		Start all instances in job array by ID
--job-array-name		string		Start all instances in job array by name

spawn status

spawn status <instance-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--check-complete		bool		Check completion status and exit with standardized codes (0=complete, 1=failed, 2=running, 3=error)

spawn stop

Stop a running instance (preserves EBS volumes)

spawn stop [instance-id-or-name] [flags]

Flags:

Flag	Short	Type	Default	Description
--job-array-id		string		Stop all instances in job array by ID
--job-array-name		string		Stop all instances in job array by name
--yes	-y	bool		Skip the confirmation prompt

spawn sweep

Manage parameter sweeps: list, check status, cancel, resume, and collect results.

Examples: spawn sweep list spawn sweep status sweep-20260116-abc123 spawn sweep cancel sweep-20260116-abc123 spawn sweep resume sweep-20260116-abc123 --max-concurrent 5 spawn sweep collect sweep-20260116-abc123 --output results.json

spawn sweep

spawn sweep cancel

Cancel a running parameter sweep and terminate its instances

spawn sweep cancel <sweep-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn sweep collect

Download and aggregate results from a completed sweep

spawn sweep collect <sweep-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--best		int		Show only top N results by metric (0 = all)
--format		string	json	Output format: json, csv, jsonl
--metric		string		Metric to rank results by (e.g. accuracy, loss)
--output-file	-f	string	results.json	Output file path
--regions	-r	stringSlice		Regions to collect from (comma-separated or repeated)
--s3-prefix		string		Custom S3 prefix for results (default: auto-detect)

spawn sweep list

List parameter sweeps

spawn sweep list [flags]

Flags:

Flag	Short	Type	Default	Description
--last		int	20	Show last N sweeps
--region		string		Filter by region
--since		string		Show sweeps created after date (YYYY-MM-DD)
--status		string		Filter by status (RUNNING, COMPLETED, FAILED, CANCELLED)

spawn sweep resume

Resume an interrupted parameter sweep from checkpoint

spawn sweep resume <sweep-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--detach		bool		Run sweep orchestration in Lambda
--max-concurrent		int		Override max concurrent instances (0 = use original)

spawn sweep status

Show parameter sweep status and progress

spawn sweep status <sweep-id> [flags]

Flags:

Flag	Short	Type	Default	Description
--check-complete		bool		Exit with standardized codes: 0=complete 1=failed 2=running 3=error

spawn team

Create and manage teams for sharing spawn instances, sweeps, and autoscale groups

spawn team

spawn team add

Add a member to a team (owner only)

spawn team add <team_id> <iam_arn>

spawn team create

Create a new team

spawn team create [flags]

Flags:

Flag	Short	Type	Default	Description
--description		string		Team description
--name		string		Team name (required)

spawn team delete

Delete a team and all memberships (owner only)

spawn team delete <team_id> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn team list

List teams you own or belong to

spawn team list

spawn team remove

Remove a member from a team (owner only)

spawn team remove <team_id> <iam_arn> [flags]

Flags:

Flag	Short	Type	Default	Description
--yes	-y	bool		Skip the confirmation prompt

spawn team show

Show team details and member list

spawn team show <team_id>

spawn terminate

Permanently terminate an instance. This is irreversible: the instance is destroyed and any non-persisted volumes are deleted. Use stop or hibernate to keep EBS volumes.

Terminate a single instance by ID or name, or an entire job array:

spawn terminate i-0abc123 spawn terminate my-instance --yes spawn terminate --job-array-name training

spawn terminate [instance-id-or-name] [flags]

Flags:

Flag	Short	Type	Default	Description
--job-array-id		string		Terminate all instances in job array by ID
--job-array-name		string		Terminate all instances in job array by name
--yes	-y	bool		Skip the confirmation prompt

spawn upgrade-spored

Replace the spored lifecycle agent on a running instance with a newer release WITHOUT terminating/relaunching the instance and without losing spored's lifecycle state (the TTL deadline, accumulated compute-seconds, and the completion / pre-stop / idle / FSx config all live in EC2 tags that the new spored re-reads on boot).

The default target is the latest released version; pin one with --version. A downgrade is refused unless --force is given. The swap is driven over SSM (keyless — works on private-subnet / no-public-IP instances), so the instance must have the SSM agent online and an instance profile (the spored role attaches AmazonSSMManagedInstanceCore, so spawn-launched instances already qualify).

The TTL deadline is absolute and tag-stored, so it is NOT reset by the restart — an instance mid-life keeps its original termination time. Linux only for now (Windows spored upgrade is a follow-up, #234).

spawn upgrade-spored <instance-id|name> [flags]

Flags:

Flag	Short	Type	Default	Description
--allow-downgrade		bool		Allow a downgrade (target older than the running version)
--timeout		duration	5m0s	How long to wait for the on-instance upgrade to complete
--version		string		Target spored version (e.g. 0.64.0); default: latest release
--yes	-y	bool		Skip the confirmation prompt

spawn validate

Validate spawn instances and configuration against compliance controls.

This command can validate:

• Running instances against compliance controls (NIST 800-171, NIST 800-53)
• Infrastructure resources (DynamoDB, S3, Lambda, CloudWatch)
• Launch configuration before launching instances

Examples:

Validate all running instances against NIST 800-171

spawn validate --nist-800-171

Validate specific instance

spawn validate --instance-id i-0abc123 --nist-800-171

Validate infrastructure resources

spawn validate --infrastructure

Output as JSON for automation

spawn validate --nist-800-171 -o json

spawn validate [flags]

Flags:

Flag	Short	Type	Default	Description
--infrastructure		bool		Validate infrastructure resources (DynamoDB, S3, Lambda)
--instance-id		string		Specific instance ID to validate
--nist-800-171		string		Validate NIST 800-171 compliance
--nist-800-53		string		Validate NIST 800-53 compliance (low, moderate, high)
--region		string		AWS region to validate (default: all regions)

spawn version

Display version, build date, and git commit information for spawn.

spawn version