Overview of agent configuration tabs

The Overview tab is where you define what the agent is, what it does, and what inputs users must provide.

This tab affects the user experience at the start of a run. It determines how the agent is presented on the run page and what information users must provide before the agent begins.

On this tab, you can:

The Parameters tab is where you define structured options that guide how the agent generates output.

This tab affects how the agent adapts its responses. It allows users to select options such as language, audience, or region, which influence tone, style, and content generation.

On this tab, you can:

The Schemas tab is where you define the structure of the agent’s output when you need consistent, predictable results.

This tab affects how results are organized and returned. It ensures the agent produces consistent, structured output that can be validated and reused.

Schemas can help fine-tune agent output. Clearly describing each field guides the model to generate more consistent and predictable responses. If structured JSON is not required, you can instead guide the output by configuring the action's system prompt (in the Workflow editor).

For example, a LinkedIn post schema might define fields like body, hashtags, headline, mentions and call_to_action, ensuring every output follows the same structure.

Schemas can be used alongside parameters you defined or when configuring actions that require a specific output format.

On this tab, you can:

{
  "type": "object",
  "properties": {
    "body": {
      "type": "string",
      "description": "Main content (max 3000 characters)"
    },
    "hashtags": {
      "type": "array",
      "description": "Relevant hashtags (3-5 recommended)",
      "items": {
        "type": "string"
      }
    },
    "headline": {
      "type": "string",
      "description": "Attention-grabbing headline or opening line"
    },
    "mentions": {
      "type": "array",
      "description": "People or companies to tag",
      "items": {
        "type": "string"
      }
    },
    "call_to_action": {
      "type": "string",
      "description": "Clear call-to-action"
    }
  },
  "required": [
    "body",
    "hashtags",
    "headline",
    "mentions",
    "call_to_action"
  ],
  "additionalProperties": false
}

The HTML Templates tab is where you define how the agent’s output is presented.

This tab affects how results are displayed. It takes structured output (defined by a schema) and formats it into a layout, such as a report or document. Templates use Handlebars syntax to map schema fields into HTML using placeholders such as {{title}} or {{body}}.

HTML templates can be used alongside parameters you defined or when configuring actions that require a specific output layout.

On this tab, you can:

<div class="linkedin-post">
  <div class="headline"><strong>{{headline}}</strong></div>
  <div class="body">{{body}}</div>
  <div class="hashtags">{{#each hashtags}}<span class="hashtag">#{{this}}</span> {{/each}}</div>
  <div class="cta">{{call_to_action}}</div>
  {{#if mentions}}<div class="mentions">{{#each mentions}}<span class="mention">@{{this}}</span> {{/each}}</div>{{/if}}
</div>

The Workflow tab is where you define the sequence of steps the agent follows to complete a task.

This tab affects how the agent processes inputs, applies logic, and generates outputs. It controls how data flows between steps and how results are created and saved.

Using the Workflow editor, you design the sequence of actions the agent executes.

The Workflow editor includes the following areas:

After building your workflow, test it by clicking Run on the canvas. This executes the workflow using the current configuration and inputs. As the workflow runs, each step is executed in sequence, allowing you to verify how data flows between actions and how outputs are generated.

Use the Runs tab in the right pane to review the results of each run. You can inspect individual steps, check inputs and outputs, and identify any issues or errors to help with debugging.

Prompts can be used in several parts of an agent configuration, including instructions, message templates, and system prompts. These prompts guide how the agent interprets inputs, applies rules, and generates outputs.

Follow these best practices to improve the reliability and usefulness of generated outputs: