Agent Studio, evolved from Vertex AI Studio, is Google’s interactive workspace for generative AI: prompt design, model tuning, and prototyping in a single interface.
What is Agent Studio?
Agent Studio is the central workspace for generative AI and agent development on the Gemini Enterprise Agent Platform (formerly Vertex AI). The interface enables developers and data scientists to test prompts, evaluate models, create GenAI prototypes, and design agents without necessarily writing code.
Agent Studio provides direct access to Google’s current Gemini model family for text, image, and other modalities, as well as additional foundation models from the platform’s Model Garden. Older, separately marketed model lines such as PaLM 2 or Codey have been discontinued in favor of Gemini models. In Agent Studio, you can iteratively refine prompts, compare different models, and export results directly to API calls or agent configurations.
For advanced use cases, Agent Studio supports supervised fine-tuning and distillation. You can train models on your specific data and create more compact, cost-effective versions. For grounding on enterprise data, Agent Studio can be combined with RAG Engine and Agent Search, and for production agent operation with Deployments/Agent Runtime.
Core Features
- Interactive prompt design and testing with direct model comparison
- Access to the current Gemini model family and additional foundation models
- Multimodal inputs (text, image, video)
- Supervised fine-tuning and distillation for custom models
- Integration with RAG Engine and Agent Search for grounding
- Export of prompts and agent configurations for production use
Typical Use Cases
Prompt Engineering and Optimization
A development team tests different prompt variants for a customer service bot. In Agent Studio, system instructions, few-shot examples, and other parameters can be adjusted iteratively while comparing outputs.
Model Evaluation and Benchmarking
A data science team evaluates different Gemini model variants for a classification task. Agent Studio enables systematic testing with defined test datasets.
Supervised Fine-Tuning
An e-commerce company trains a Gemini model on product descriptions. Using its own example data, Agent Studio produces a specialized model for consistent product texts.
Multimodal Prototypes
An insurance company tests damage image analysis with a multimodal Gemini model. In Agent Studio, sample images are uploaded and prompts are developed that detect and categorize damage.
Agent Prototyping
A business unit designs an initial agent prototype in Agent Studio, grounded on internal documents via Agent Search, before moving the agent to the managed runtime for production use.
Integration with innFactory
As a certified Google Cloud partner, innFactory supports you in developing GenAI applications: from prompt engineering through fine-tuning to production architecture on the Gemini Enterprise Agent Platform.
Contact us for an AI consultation.
Available Tiers & Options
Prompt and Agent Design
- Interactive prompt testing and comparison
- Current Gemini models available
- Multimodal inputs (text, image, video)
- Production use requires the API/runtime in addition
Tuning
- Supervised fine-tuning
- Distillation
- Requires your own training data
Typical Use Cases
Technical Specifications
Frequently Asked Questions
What is Agent Studio (formerly Vertex AI Studio)?
Agent Studio is Google's interactive workspace for generative AI and agents within the Gemini Enterprise Agent Platform. It evolved from the former Vertex AI Studio and provides an interface for testing prompts, fine-tuning models, and prototyping GenAI applications and agents.
Which models are available in Agent Studio?
Agent Studio provides access to the current Gemini model family for text, image, and other modalities, plus additional foundation models from the platform's Model Garden. Older, separately marketed model lines such as PaLM 2 or Codey have been discontinued in favor of Gemini models.
What is the difference between Agent Studio and using the API directly?
Agent Studio is the visual environment for exploration, prompt design, and prototyping. Production workloads use the corresponding Gemini Enterprise Agent Platform API directly, or are deployed onto the managed runtime (Deployments/Agent Runtime). Prompts and agent configurations created in Agent Studio can be exported.
How much does using Agent Studio cost?
The interface itself does not carry a separate base cost; billing is usage-based for underlying model usage (input/output tokens), training, and tuning. Exact prices are available in the official Google Cloud pricing list.
Can I use my own models in Agent Studio?
Yes, you can use fine-tuned versions of the available foundation models or bring your own models via the platform's Model Garden. Agent Studio supports testing the models available within the Gemini Enterprise Agent Platform.
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.
