gcloud-colab

name	gcloud-colab
description	Colab Enterprise via gcloud (`gcloud colab`). Manage Colab Enterprise resources — executions, runtime-templates, runtimes, schedules.

gcloud colab — Colab Enterprise

Overview

gcloud colab manages Colab Enterprise, a managed, collaborative notebook environment that runs Jupyter/Colab notebooks on Google Cloud with enterprise security and IAM. Its API is a subset of the Vertex AI API. Reach for it to provision notebook compute (runtimes) from reusable VM blueprints (runtime templates), and to run notebooks headlessly as one-off executions or on recurring cron schedules. All resources are regional, so most commands take --region (or default it with gcloud config set colab/region).

Quick reference — common workflows

1. Enable APIs and set a default region

gcloud services enable aiplatform.googleapis.com dataform.googleapis.com compute.googleapis.com
gcloud config set colab/region us-central1

2. Create a runtime template (the VM blueprint)

# Minimal template (defaults: e2-standard-4, 100 GB PD_STANDARD disk)
gcloud colab runtime-templates create \
    --display-name=my-runtime-template \
    --region=us-central1

# Template with a GPU accelerator
gcloud colab runtime-templates create \
    --display-name=my-gpu-template \
    --machine-type=n1-standard-2 \
    --accelerator-type=NVIDIA_TESLA_V100 \
    --accelerator-count=1 \
    --region=us-central1

gcloud colab runtime-templates list --region=us-central1

3. Create and manage a runtime (the running VM)

gcloud colab runtimes create \
    --display-name=my-runtime \
    --runtime-template=my-runtime-template \
    --region=us-central1

gcloud colab runtimes stop my-runtime --region=us-central1     # stop to save cost
gcloud colab runtimes start my-runtime --region=us-central1    # restart
gcloud colab runtimes upgrade my-runtime --region=us-central1  # upgrade the image
gcloud colab runtimes delete my-runtime --region=us-central1

4. Run a one-off notebook execution (headless)

gcloud colab executions create \
    --display-name=my-execution \
    --notebook-runtime-template=my-runtime-template \
    --gcs-notebook-uri=gs://my-bucket/my-notebook.ipynb \
    --service-account=[e-mail] \
    --gcs-output-uri=gs://my-bucket/results \
    --region=us-central1

gcloud colab executions describe my-execution --region=us-central1
gcloud colab executions list --region=us-central1

5. Schedule a recurring notebook run

gcloud colab schedules create \
    --display-name=my-schedule \
    --cron-schedule='TZ=America/Los_Angeles 0 8 * * *' \
    --max-concurrent-runs=1 \
    --start-time=2026-07-01T00:00:00Z \
    --execution-display-name=my-execution \
    --notebook-runtime-template=my-runtime-template \
    --gcs-notebook-uri=gs://my-bucket/my-notebook.ipynb \
    --service-account=[e-mail] \
    --gcs-output-uri=gs://my-bucket/results \
    --region=us-central1

gcloud colab schedules pause my-schedule --region=us-central1
gcloud colab schedules resume my-schedule --region=us-central1
gcloud colab schedules update my-schedule \
    --cron-schedule='TZ=America/Los_Angeles 0 9 * * *' \
    --max-runs=10 \
    --region=us-central1

6. Grant per-template access to a user

gcloud colab runtime-templates add-iam-policy-binding my-runtime-template \
    --region=us-central1 \
    --member=user:[e-mail] \
    --role=roles/aiplatform.notebookRuntimeUser

gcloud colab runtime-templates get-iam-policy my-runtime-template --region=us-central1

Command groups

Group	Reference file	Commands	Description
colab executions	executions.md	4	manage Colab Enterprise notebook execution jobs
colab runtime-templates	runtime-templates.md	8	manage Colab Enterprise runtime templates
colab runtimes	runtimes.md	7	manage Colab Enterprise runtimes
colab schedules	schedules.md	7	manage Colab Enterprise notebook execution schedules

See index.md for a one-line index of all 26 commands.

Common flags & tips

• Region is required (or defaulted). Most commands take --region; set a default with gcloud config set colab/region us-central1 to omit it. The colab/region property backs every resource's region attribute.
• Resource identity. Templates, runtimes, executions, and schedules are addressed by ID. For runtime-templates create and runtimes create you may pin the ID with --runtime-template-id / --runtime-id; otherwise a random ID is generated.
• Notebook source (executions & schedules). Provide the notebook one way: --gcs-notebook-uri (Cloud Storage, optionally --generation), a Dataform repo (--dataform-repository-name with --commit-sha), or --direct-content (executions only). Pair it with --gcs-output-uri for results and an identity (--service-account or --user-email).
• Execution timeout. --execution-timeout defaults to 24h; accepts duration strings (see gcloud topic datetimes).
• Runtime template sizing & security. --machine-type (default e2-standard-4), --disk-size-gb (100), --disk-type (PD_STANDARD), --accelerator-type/--accelerator-count, --idle-shutdown-timeout (3h), plus --enable-secure-boot, --no-enable-euc (disable end-user credentials), --no-enable-internet-access, --network/--subnetwork, --network-tags, and CMEK via --kms-key.
• Long-running ops. create/delete/start/stop/upgrade accept --async to return immediately.
• Cron timezones. Prefix the cron string with TZ= or CRON_TZ= plus an IANA zone, e.g. 'TZ=America/Los_Angeles 0 8 * * *'. Schedules end at --end-time or after --max-runs; control concurrency with --max-concurrent-runs and queueing with --enable-queueing. schedules resume can backfill missed runs with --enable-catch-up.
• Listing. list commands support --filter, --limit, --page-size, --sort-by, and --uri. Example: gcloud colab runtimes list --region=us-central1 --filter="displayName:my-*" --sort-by=createTime.
• Template IAM. Grant per-template access with add-iam-policy-binding/remove-iam-policy-binding (--member, --role), inspect with get-iam-policy, or apply a file with set-iam-policy POLICY_FILE. roles/aiplatform.notebookRuntimeUser lets a principal create runtimes from that template.

beta / alpha

gcloud beta colab mirrors the GA surface — the same four subgroups (executions, runtime-templates, runtimes, schedules) with no documented beta-only commands. No gcloud alpha colab surface exists. Prefer GA (gcloud colab) for all documented commands.

Official documentation

• Colab Enterprise docs home — getting started, guides, and resource concepts.
• Introduction — what Colab Enterprise is and how it relates to Vertex AI.
• Quickstart — enable APIs, create a notebook, run code, manage notebooks.
• Access control (IAM) — Colab Enterprise admin/user roles and underlying Vertex AI roles.
• Create a runtime template — machine type, accelerators, disk, networking, and security options.
• Manage runtimes — start, stop, disconnect, reconnect, and delete runtimes.
• gcloud colab CLI reference — full command/flag reference for all four groups.