Design agents in Agent Studio

Preview

The design agents feature in Agent Studio is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the Service Specific Terms. Pre-GA features are available "as is" and might have limited support. For more information, see the launch stage descriptions.

Relationship to Other Terms. The design agents feature in Agent Studio helps Customers design "AI Agents" as defined in the Service Specific Terms and thus the "Agentic AI Services" Service Specific Terms apply. When using Google Search as a tool in Agent Studio, the section entitled "Grounding with Google Search" in the Service Specific Terms also applies, except that the first reference to 'Gemini Enterprise or Agentspace' in subsection (iv) of the "Agentic AI Services" section is replaced with 'Gemini Enterprise Agent Platform'. Please ensure you review the terms of any other tools used in Agent Studio and verify that your agent is appropriate for deployment in Agent Runtime before utilizing that feature.

This page provides an overview of how to use Agent Studio in the Google Cloud console.

Agent Studio is a low-code visual designer within the Google Cloud console that simplifies agent development. You can visually map agent workflows, test responses in real time, and experiment with different configurations before deploying or transitioning to code.

This document provides an overview of Agent Studio and explains how to set up your environment, create and test agents, and deploy them directly to a production runtime.

Set up your environment

Before using Agent Studio, get set up with Google Cloud.

Get the required roles

To get the permissions that you need to use Agent Studio, ask your administrator to grant you the Agent Platform User (roles/aiplatform.user) IAM role on your project. For more information about granting roles, see Manage access to projects, folders, and organizations.

You might also be able to get the required permissions through custom roles or other predefined roles.

Create an agent

Use the following steps to design and test an agent in Agent Studio:

1. In the Google Cloud console, go to the Agents page.
   

   Go to Agents

2. Click Create agent to open the Agent Studio canvas for a new agent.

3. Design your agent in the Agent Studio canvas. You can switch between the following tabs:

   • Flow: Create the main agent and subagents by using a visual representation of your agent's workflow and control logic.

     1. Click an agent to open the Details panel for that agent. You can also click Add a subagent (+) to add subagents.

     2. Configure your main agent and subagents in the Details panel:

        1. Name: Add a name to help identify the agent.

        2. Description: A summary of your agent's purpose.

        3. Instructions: Add instructions to guide your agent.

        4. Model: Select the model to power your agent.

        5. Tools: Click Add tools (+) to add tools that let the agent complete tasks. For more information, see Set up and add tools.

   • Preview: Test the agent's capabilities and response as you build your agent. Chat with your agent to test its capabilities.

4. Click Get code to see your agent code. If you want to continue developing your agent elsewhere, you can copy the code and paste it to a code editor of your choice. See ADK tutorials for more options to continue developing your agent.

Once your agent is complete, you can deploy it directly from Agent Studio. For more information, see Deploy an Agent from Agent Studio.

Set up and add tools in Agent Studio

You can configure the following tools for your agent:

• Google Search: Lets the agent perform web searches using Google Search. Toggled on by default.

• URL context: Let the model analyze URLs from prompts sent to the agent. Toggled on by default.

• Agent Platform Search Data Store: Click Add (+) to let your agent access information that has been indexed in your Vertex AI Search data store.

  • Project Number: The Google Cloud project number associated with your Vertex AI Search data store. View your project number.

  • Location: The location of your data store.

  • Data Store ID: Firestore in Datastore mode (Datastore) ID of the data to include. View a list of your data stores and IDs.

  • Collection ID: Collection ID of the data to include. View a list of your data stores and collection IDs. If your data store doesn't have a Collection ID, enter default_collection.

  If you don't have an existing data store, see Get started with custom search to create one. Then grant service account access to Agent Platform Search:

  1. In the Google Cloud console, go to the IAM page.
     

     Go to Identity and Access Management (IAM)

  2. Click Grant access.

  3. In the New principals field, enter the following service account information: [e-mail].

  4. Select + Add roles. Search for and select Discovery Engine User. Click Apply and then Save.

• MCP Server: Click Add (+) to add MCP tools by connecting to an MCP server.

  1. MCP display name: Enter a name for your MCP server.

  2. Endpoint URL: Enter an endpoint URL for the MCP server.

  3. Authentication: Autofilled as None. Agent Studio only supports MCP servers that don't require authentication.

  Your agent can use all tools in your connected MCP server.

Deploy an Agent from Agent Studio

Once you create and preview an agent, you can deploy it to production. Use the following steps to deploy an agent from Agent Studio:

1. From the Agents list page, click the agent you want to deploy. The Agent detail page appears for the selected agent.

2. Click Deploy to open the Deploy to an Agent Runtime instance dialog.

3. (Optional) Edit the Display name or Description for your agent.

4. Select a deployment region from the list of available regions, then click OK.

5. Click Deploy.

Deployment creates a new runtime instance and can take up to five minutes to complete. Upon success, a message appears on the Flow tab of the Agent Studio canvas. Your agent is now available in production and can integrate securely with your external apps.

View a deployed agent

To view your deployed agent:

1. In the Google Cloud console, go to the Agent Platform page.
   

   Go to Agent Platform

2. Use the Region list to filter by deployment region.

3. Deployed agents that are part of the selected project appear in the list.

4. Click the name of the specified agent. The Metrics page for the agent opens.

For more information on available metrics for your agent, see View metrics for your deployed agent.