C3 AI Documentation Home

Configure Agents

Use this guide to configure an agent in the by defining its objectives, prompts, model selection, memory, and tools so it performs tasks consistently and responds as intended.

Prerequisites

Before configuring an agent, ensure the following:

  • You have created an agent in the Agent Gallery. If you have not yet created an agent, see Create Agents.
  • You have the required role or permissions to edit agent configuration. Learn more about the roles in Agent and Tool Gallery

Configure an agent through the Agent Workbench

The Agent Workbench displays as a split-screen interface with three panels:

  • Configuration (left): Controls for objectives, prompts, model selection, memory, and tools.
  • Chat (center): Live chat interface to test agent responses.
  • Trace (right): Execution traces for debugging and monitoring.

All panels are collapsible. The Trace panel is collapsed by default.

For details on the Agent Workbench, see Agent Workbench.

GenAI Workbench Preview

Follow these steps to configure an agent:

  1. Define your agent's purpose in the Objective and Prompts section. Write clear, specific instructions that explain what your agent should do and how it should go about doing them. If you are adding tools for your agent to use you should also instruct the agent on when to use which tools, and how it should use them. The clearer and more descriptive you are, the more likely your agent is to perform its required tasks as intended.

    Example:

    Text
    You are a helpful customer support agent. When customers ask questions about our products, provide clear, friendly answers. If you don't know something, politely say so and suggest they contact a human agent for more help.

  2. Select the LLM Client in the Model section.

    Available models vary by application and environment. For information on configuring LLM clients, see Configure LLM Clients.

  3. Set up conversation memory using the Short Term Memory and Long Term Memory toggles. Short term memory retains context within a single conversation, while long term memory retains information across multiple interactions. For most applications, enable this feature to provide a better user experience.

  4. If you are configuring a Dynamic agent, use the Data Model section when your agent needs structured queries over application data. Create a data model, select data sources and fields, and generate Data Model Documentation. For step-by-step guidance, see Configure Data Models in Agent Workbench in the genAiSearch documentation (genai-dev-configure-agent-data-model.c3doc.md).

  5. Add tools to extend your agent's capabilities. Select the plus icon (+) in the Tools section to load the Add Tools modal. Here you you will get a list of tools that can you select from and add to your agent. At the top left of the modal, you can also search for specific tools by name or filter tools by specific tags. At the top right of the modal, you can click + Create Tool to create a new tool.

    To create new tools, see:

  6. Add skills to provide your agent with reusable instructions and resources for specific tasks. Select the plus icon (+) in the Skills section to browse and add available skills. To learn more on how to create skills, see Create and Add Skills.

Add Documentation

Agent documentation helps other users understand your agent's purpose and capabilities.

  1. Select the Documentation button in the upper right area of the workbench.
  2. Select the pencil icon to enter edit mode.
  3. Describe what your agent does, what tools it uses, and any limitations.
  4. Provide example interactions to demonstrate expected behavior.

Select Save regularly throughout the configuration process to preserve your work.

Once your agent is configured and tested, it is ready to be deployed. Continue with Deploy Agents to make your agent available for use.

See also

Was this page helpful?