Creating an Agent Using the Form

How to Create an Agent using the Guided Form

Kadal allows users to create AI-powered agents to automate tasks using simple prompts or structured forms. Agents can process documents, follow guidelines, and generate contextual responses, making your work faster and more accurate. Kadal’s Guided Form View lets you manually set up AI agents with complete control over configuration, files, and model preferences.

The Agents page contains two tabs:

1. Published: Displays agents that have been published by any user within the tenant.
2. Drafts: Display agents that are still in development or undergoing trials before being published.

Note: All newly created agents are stored in Drafts until published. If a user wants to share an agent for others to use, they can publish it. Published agents can be unpublished by the agent's creator or an admin. Once unpublished, the agent can be edited.

Steps:

1. Log in to Kadal
   • Access your Kadal account with your credentials.
   • Navigate to the "Try and Build" page.
2. Start Building the Agent
   • Click on the Agent card to enter the agent creation section.
   • In the top-right corner, click "+ Build Your Agent" to begin the process.
3. Configure Agent Details
   • Agent Name: Provide a unique and descriptive name to identify the agent.
   • Description: Write a brief overview outlining the agent’s purpose and functionality.
   • Master Prompt: Define the core instructions that guide the agent’s responses.
4. Add Agent Sources: You can attach reference material that the agent will use to generate accurate and context-specific responses. Kadal offers three types of sources:
   • Content: Add files and folders from the repository.
   • Guidelines: Upload external files and folders directly from your system.
   • Templates: Upload predefined templates from your system.
   • To add sources:
     • Click +Add to attach content as the agent's knowledge base.
     • In the popup, select the required files from the repository and click Add.

Note: Users can only select files available in the repository. Ensure that the necessary files are added to the repository before starting the agent creation process.

5. Configure Advanced Settings
   • Model: Choose from the available AI models (e.g., GPT, Gemini, Claude).
   • Response Type: Adjust the randomness of responses (lower values increase accuracy, while higher values add variation).
   • Max Tokens: Define the maximum output length (1 token ≈ 4 characters).

Note: Kadal supports multiple AI models for Agent creation, some are default, while others can be configured upon request.

6. Save and Draft the Agent
   • Click “Save” to store the agent configuration.
   • The agent will be saved in the Drafts section.
   • You can run test trials before publishing the agent.

For detailed steps on editing a draft agent, refer to our guide on Editing a Draft Agent.

Best Practices

• Use Clear and Concise Prompts: Ensure the Master Prompt is specific and unambiguous to generate accurate responses.
• Select Relevant Sources: Add well-organized content and templates to improve the agent’s contextual accuracy.
• Optimize Response Settings: Fine-tune the response type and token limit for the desired output quality.
• Select an AI model that best aligns with your agent’s purpose (e.g., GPT for general tasks, Claude for structured outputs).
• Test Before Publishing: Run multiple test trials in the Drafts section to verify the agent’s accuracy and performance.

By following these steps, you can build and configure a custom AI agent in Kadal, ready for seamless interaction and task execution.