AI agents can automate repetitive tasks, improve productivity, and help teams work more efficiently. With Gemini’s Agent Designer Builder, you can create custom agents that handle specific business processes using a visual interface, no complex development required.
In this guide, we’ll walk through the process of building an agent using Agent Designer Builder and demonstrate how agents and subagents can work together to complete multi-step workflows.
Why Use Agent Designer Builder?
Agent Designer Builder allows organizations to create AI assistants that can:
- Automate routine tasks
- Standardize business processes
- Route work between specialized agents
- Reduce manual effort
- Improve response quality and consistency
One of the platform’s most powerful features is the ability to create subagents that handle specific responsibilities within a larger workflow.
Step 1: Create a New Agent Application
- Log into the Google Cloud Console and navigate to the Gemini Enterprise Agent Platform (formerly Vertex AI Agent Builder).
- In the left-hand navigation menu, click on Agent Studio (the low-code visual designer environment).
- Click the + New Agent or Create App button.
- Select Proceed to Builder to launch the visual, node-based workspace interface.
- You will see the Default Start Node on your canvas, which acts as the entry point for your conversational workflow.
Step 2: Configure the Main Agent Node
- Click on the starting main agent node on the canvas to open its configuration panel on the right.
- Customize the primary attributes:
- Agent Name: Give your root agent a clear identifier.
- Model Selection: Choose the foundational model powering the reasoning engine (e.g., Gemini 2.5 Flash or Gemini 2.5 Pro). Learn more about model capabilities in the Google Cloud Vertex AI Documentation.
- Goal / Description: Briefly define what this root agent is meant to accomplish.
- Instructions: This is your system prompt. Explicitly define the main agent’s persona, boundaries, and how it should initially greet or handle user input.
Step 3: Add and Link a Subagent
The platform supports a multi-agent framework where the main agent can hand off specialized tasks to subagents.
- Click back onto the visual builder canvas.
- Hover over the main agent node, or right-click the canvas, and click the + (Add Node) icon.
- Select Subagent.
- Connect the main node to this new node via a visual connector line to establish the routing flow.
- Click the new subagent node to configure its isolated parameters:
- Name & Specialized Instructions: Give it strict parameters. (e.g., If the main agent handles general customer triage, this subagent might have instructions exclusively for Processing Refunds or Searching Technical Data Stores).
Step 4: Define Routing and Handoff Instructions
For a multi-agent system to work smoothly without getting stuck in loops, you must explicitly program the transition rules.
- In the Main Agent’s Instructions, include specific execution conditions.Example: “If the user asks a question regarding data analysis, immediately route the conversation to the Analytics Subagent.”
- In the Subagent’s Instructions, define the exit criteria.Example: “Once the refund is processed or if the user asks an unrelated question, hand control back to the Main Agent.”
- Connect any necessary Tools (like OpenAPI extensions) or Data Stores (Vertex AI Search) directly to the specific subagent that requires them.
Step 5: Test and Simulate the Multi-Agent Flow
- Look to the right side of the screen to find the Preview / Simulator panel (or click Chat with Agent).
- Start a test conversation to evaluate the execution graph in real time.
- Verify the following during testing:
- Does the main agent respond using the correct tone and constraints?
- When you trigger a specific sub-task, does the visual graph show the conversation successfully handing off to the subagent?
- Does the subagent have proper access to its assigned tools or data stores?
- If an agent fails to route or breaks character, click into that specific node, adjust the text in the Instructions panel, and click Save/Train to test again.
Pin Your Agent for Easy Access
After testing your agent:
- Navigate back to the Agents section.
- Locate the agent under Your agents.
- Click the three-dot menu on the agent card.
- Select the Pin option.
- The agent will now appear in the left-hand navigation menu, making it easy to access whenever needed.
Unlock the Full Potential of Gemini Enterprise
Custom agents are one of the most powerful capabilities available in Gemini Enterprise. From sales and customer service to operations and project management, organizations can build agents that automate workflows, improve efficiency, and help teams focus on higher-value work.
Whether you’re building your first AI agent or developing a broader AI transformation strategy, HiView Solutions can help you maximize the value of Gemini Enterprise and accelerate business results. Check out our page to learn more!
