A chat agent is a type of agent that you can build using Workflow Builder in your Gemini Enterprise web app. These agents are designed to interact with users in a conversational manner to complete a specific task. Chat agent responses depend on the user's input, making them primarily non-deterministic in nature.
Chat agents are ideal for the following use cases:
- Open-ended conversations without fixed steps. For example, a brainstorming assistant for project ideas.
- Dynamic tasks where user input drives the flow. For example, a customer support agent where the next question depends on the user's issue.
- Q&A, data lookup, or personalized assistance. For example, checking Jira ticket status or summarizing a specific email thread.
Before you begin
Before exploring these features, do the following:
- Ensure that a Gemini Enterprise administrator has enabled the Enable chat agents feature toggle in your app's feature management settings. For more information, see Manage web app features.
- Ensure you have a Gemini Enterprise web app with the necessary data stores connected. For more information, see About apps and data stores.
Chat agent canvas
The chat agent canvas is an interactive interface that you can use to build and refine your chat agent. The canvas is divided into two main modes:
Chat mode: the chat mode provides a conversational interface to build and refine your agent using natural language prompts. It's perfect for no-code users who want to use Gemini's intelligence to quickly develop and adjust their agent's details, instructions, or behaviors.
Designer mode: the designer mode provides a visual representation of your agent and is organized into two panes:
- Chat pane: this pane is similar to the chat mode, where you can interact with your agent in a conversational manner while viewing the changes to the agent in the designer pane.
Designer pane: this pane provides a visual representation of your agent and is further divided into the following tabs:
- Flow: presents a visual representation of your agent's workflow and control logic. Use the Flow tab to view the overall details of your main agent, manage the main agent and all individual sub-agents, and to design complex, multi-step agents.
- Schedule: provides configuration options to set one or more execution schedules for your agent.
- Preview: provides a live, interactive test environment for your agent. Use the Preview tab to test the agent's functionality and response as you build.
Chat agent components
Chat agents can be single-step or multi-step agents. Single-step chat agents consist of a single agent node, which contains the logic for the agent. Multi-step chat agents are made up of a main agent node that is connected to one or more sub-agent nodes. Each sub-agent node is an independent agent with its own logic.
What's next
- Learn how to create a chat agent.
- Learn how to create a workflow.
- Learn how to use a chat agent.