Skip to main content

​
Why Transform Responses?

  • Token efficiency: Summarizes to the few facts the model needs next, conserving context window and reducing distraction.
  • Better tool use loops: Framing outputs as intentions (“Inventory insufficient; check warehouse B”) helps the LLM choose the right next tool call.
  • Schema abstraction: Hides brittle field names and shapes, so prompts don’t break when an endpoint version changes.
  • Consistent tone & policy: Centralized phrasing enforces brand, compliance language, and safety policies across all tool outputs.
  • Disambiguation of codes & enums: status=3 → Payment declined because the card expired, which the model can reason about directly.

​
How Response Formatting Works

1

API Response Receipt

Raw HTTP response received from your API endpoint
2

Template Processing

Handlebars template processes the context to generate formatted output
3

Response Delivered to AI Agent

Structured response delivered to AI agent for natural language understanding
Response formatting flow from API response to AI agent output

​
Formatting Configuration System

​
Core Configuration Options

  • Main Template
  • Collection Template
{{! Primary template for formatting entire response }}
**User Profile Retrieved**

Name: {{response.data.body.full_name}}
Email: {{response.data.body.email}}
Status: {{response.data.body.account_status}}

{{#if response.data.body.preferences}}
**Preferences:**
{{#each response.data.body.preferences}}
- {{@key}}: {{this}}
{{/each}}
{{/if}}

​
Template Variables and Context

​
Available Variables

The formatting system provides a rich context for template processing:
  • Response Data
  • Request Context
  • Error Context
  • Loop Variables
{{! Response object structure }}
{{response.data.body}}           <!-- Main response data -->
{{response.data.headers}}        <!-- Response headers -->
{{response.data.status}}         <!-- HTTP status code -->

{{! JSON stringify entire response }}
{{json response}}

{{! Access nested properties }}
{{response.data.body.user.name}}
{{response.data.body.results.0.title}}

​
Built-in Handlebars Helpers

Converts objects to JSON strings for structured output.
{{json response.data.body}}
{{json request.data.parameters}}

<!-- Formatted JSON with indentation -->
```json
{{json response.data.body}}

</Accordion>

<Accordion title="Comparison Helpers">
Logical comparison operations for conditional rendering.

```handlebars
{{! Equality checks }}
{{#if (eq status "active")}}User is active{{/if}}
{{#if (ne role "admin")}}Limited access{{/if}}

{{! Numeric comparisons }}
{{#if (gt price 100)}}Premium product{{/if}}
{{#if (lt stock 5)}}Low inventory warning{{/if}}
Combine multiple strings or values.
{{concat "User: " user.name " (" user.role ")"}}
{{concat "API Response from " request.data.parameters.path}}
Encode strings to Base64 format.
{{base64 "sensitive data"}}
{{base64 response.data.body.token}}

​
AI-Generated Templates

AI template generation interface with Sparkles icon and generation options

​
Automatic Template Generation

AgentPass includes intelligent template generation powered by AI:
1

Analyze API Response

AI examines the API response structure and data types
2

Understand Context

AI considers the tool’s purpose and expected use cases
3

Generate Template

AI creates an optimized Handlebars template for the response
4

Optimize for Agents

Template is structured for optimal AI agent comprehension
⌘I