How to Create and Configure an Agent

Choose your Agent. Give it instructions. Launch.

Agents are the core execution tools in Ongage Studio. Creating a new one is fast, flexible, and fully guided. You don’t need to write a prompt or code a workflow; you simply configure a set of fields, assign a Brand Voice, and let the Agent handle the task.

This guide walks you through the full process of creating and configuring a new Agent from scratch.

What You’re Creating

Each Agent is a modular tool trained to complete one task, such as writing, formatting, translating, verifying, or tagging. You control its behavior through:

Input Fields: What the Agent needs to know before it runs
Defaults: Pre-filled values and examples to guide usage
Brand Voice: Optional tone and writing style applied dynamically
Advanced Logic: Prompt-level variables that shape the Agent’s output
Image Toggle (Optional): Enables image generation if supported by the Agent

Once configured, Agents can be reused, duplicated, and run by anyone with access to your Studio. Studio.

Step-by-Step: Creating a New Agent

1. Open the Agent Gallery

Go to Create Content → Agent Gallery
Click + Create New Agent
Choose a base type (e.g. Blog Post, SEO Block, Summarizer)

Each base type comes with its own internal prompt logic, supported output types, and capabilities.

Some Agent types (like Listicle or Blog Post Generators) also support image output via a preconfigured image prompt. If your selected base supports this, you'll be able to toggle it on during setup.

2. Name and Categorize the Agent

Give your Agent a clear name (e.g. Product Launch Email or Grace Harper Blog Post)
Assign it to a content category (Article, Email, Utility, etc.)
Optionally link it to a Site

Tip: Use consistent naming for Brand Voice or workflow stage (e.g. Listicle Step 1 – Weblog).

3. Configure Inputs

Every Agent requires at least one input, usually a text prompt, source, or goal.

For each field, define:

Label: What the user will see (e.g. “Article Topic”)
Type: Text, dropdown, checkbox, multi-line, date, numeric, etc.
Required: Mark as mandatory or optional
Default Text Inside Field: A pre-filled suggestion or instruction
Token Placeholder: The {{token}} that maps to the Agent’s prompt logic

Each input should clearly explain what the Agent needs to generate the best result.

4. Add Advanced Configuration (Optional)

You can include advanced options like:

Brand Voice selector
LLM model dropdown (e.g. Claude 3.7, GPT-4o, O1 Mini)
Image model dropdown (e.g. Ideogram, Imagen, Seedream)
Content formatting type (Plain text, HTML, JSON)
Static fields (e.g. {{CacheBuster}} = true to prevent stale outputs)
Hidden logic fields for system or automation use

These settings let you fine-tune execution behavior, which is especially useful for Agents used in chains or with advanced formatting rules.

Only some Agents support image generation. If your Agent includes an Image Prompt, you can activate image output using an image toggle or description token like {{context.imageDescription}}.

5. Link a Brand Voice (Optional)

Assigning a Brand Voice makes sure your Agent always writes in the right tone, structure, and style.

You can allow users to:

Choose from a list of Brand Voices
Lock the Agent to a single Brand Voice
Leave it open for manual assignment during use

6. Save and Test

Once your Agent is configured:

Click Save Agent
Run a test input to verify formatting, output quality, and Brand Voice alignment
Adjust your input logic or prompt if needed

Every run is saved in My Content, and the Agent is now available from the main dashboard or its assigned Site.

Best Practices

Keep inputs minimal: Only ask for what’s essential. More inputs = more complexity.
Use clear field labels: Treat your input fields like UI instructions.
Preview default outputs: Every Agent should generate clean, complete text on the first try.
Test for edge cases: Try short inputs, empty fields, or unusual content to make sure your Agent is resilient.
Clarify image support: Only enable image toggles for Agents with image prompts