Agent Platform Workbench is the managed notebook environment of the Gemini Enterprise Agent Platform and evolved from Vertex AI Workbench. It provides Jupyter notebook-based development environments for the entire data science workflow - from data exploration to model development.
What is Agent Platform Workbench?
Workbench provides JupyterLab instances that already come with common deep learning libraries such as TensorFlow and PyTorch. Instances can be configured CPU-only or GPU-enabled, depending on the requirements of the task at hand.
For working with data, Cloud Storage and BigQuery are integrated directly into JupyterLab; for distributed processing, there is an integration with Managed Service for Apache Spark. Notebooks can be executed once or on a recurring schedule through the scheduler, and versioning happens through GitHub repository synchronization.
For cost management, the documentation lists automatic idle shutdown with customizable timing as well as Compute Engine reservation support. On the security side, Google Cloud authentication and authorization, VPC network configuration options, customer-managed encryption keys, and Confidential Computing are available.
Customization is possible through Conda environments for additional kernels and through custom containers built on the base images Google provides. Workforce Identity Federation can be used for third-party credentials. Billing runs through the Gemini Enterprise Agent Platform price list, where platform and generative AI services are priced separately.
Core Features
- Prepackaged JupyterLab: Instances with pre-installed deep learning libraries, CPU-only or GPU-enabled.
- Data access: Cloud Storage and BigQuery inside JupyterLab, plus Managed Service for Apache Spark.
- Scheduler: One-time or recurring notebook execution.
- Cost management: Automatic idle shutdown and Compute Engine reservation support.
- Security: VPC configuration, CMEK, and Confidential Computing.
- Customization: Conda environments, custom containers, and Workforce Identity Federation.
Typical Use Cases
Exploratory analysis on BigQuery data
A data science team analyzes datasets directly from the notebook without exporting data first.
GPU-based model development
A GPU instance is used to train smaller models and released again through automatic idle shutdown once work is finished.
Recurring analyses
A notebook that calculates metrics regularly is automated through the scheduler.
Distributed processing of large datasets
For compute-intensive preprocessing, Managed Service for Apache Spark is used from within the notebook.
Reproducible working environments
A team standardizes its environment through custom containers built on Google base images and synchronizes notebooks with a GitHub repository.
Benefits
- Ready to work: Pre-installed libraries remove the need to build environments from scratch.
- Short path to data: BigQuery and Cloud Storage are integrated directly into JupyterLab.
- Cost-aware operation: Automatic idle shutdown avoids paying for idle instances.
- Enterprise-grade security: VPC configuration, CMEK, and Confidential Computing are available.
- Connected to the Agent Platform: Notebooks work directly with Gemini Enterprise Agent Platform services.
Integration with innFactory
As a certified Google Cloud Partner, innFactory supports you with Agent Platform Workbench: building standardized notebook environments, connecting to BigQuery and Cloud Storage, cost and security concepts, and migrating existing Vertex AI Workbench environments. We show how a company-wide AI stack can be built using our CompanyGPT cloud stack as an example.
Contact us for a consultation on Agent Platform Workbench and Google Cloud.
Typical Use Cases
Technical Specifications
Frequently Asked Questions
What is Agent Platform Workbench?
Agent Platform Workbench - officially Gemini Enterprise Agent Platform Workbench and formerly Vertex AI Workbench - provides, according to the documentation, Jupyter notebook-based development environments for the entire data science workflow. From within notebook instances you can work directly with the Gemini Enterprise Agent Platform and other Google Cloud services.
Which libraries are pre-installed?
Instances are prepackaged with JupyterLab and deep learning libraries such as TensorFlow and PyTorch. Additional kernels can be added through Conda environments, and custom containers can be created from the base images Google provides.
How can costs be controlled?
The documentation describes automatic idle shutdown with customizable timing. Compute Engine reservations are also supported. Billing follows the Gemini Enterprise Agent Platform price list, where platform and generative AI services are priced separately.
What security features are available?
Google Cloud authentication and authorization, VPC network configuration options, customer-managed encryption keys (CMEK), and Confidential Computing are listed. Workforce Identity Federation is available for third-party credentials.
Can I run notebooks automatically?
Yes. The documentation describes a notebook execution scheduler for one-time or recurring runs.
How does Workbench relate to the Gemini Enterprise Agent Platform?
Workbench is part of the Gemini Enterprise Agent Platform, the evolution of Vertex AI. Notebook environments access platform services directly, and billing runs through the platform price list.
Note: All product information on this page has been compiled with care, but is provided without guarantee and may be outdated or incomplete. Cloud services evolve rapidly — features, pricing, SLAs, and availability change frequently. Authoritative and up-to-date information can only be found on the official product page of Google Cloud (official documentation). This page does not represent an offer by Google Cloud.
