1) Open the toolbox
• Go to Workforce → select the Worker → Toolbox → Create tool.
────────────────────────────────────────
2) Choose the type
• Tool Type: HTTP Request.
────────────────────────────────────────
3) Fill in the basic details
• Name: short and descriptive (e.g., crm_find_client).
• Description: one sentence describing its purpose.
• URL: paste the base or full endpoint URL (https).
• Method: GET, POST, PUT, DELETE, PATCH.
• Endpoint (if your URL is split): complete the path.
• Include response: check Body and/or Status depending on what you need the AI to interpret.
────────────────────────────────────────
4) Headers (optional but common)
• Add Header → insert key/value pairs (e.g., Authorization: Bearer <TOKEN>, Content-Type: application/json).
────────────────────────────────────────
5) Body — how to create objects, arrays, and reference Inputs
• Add Body Fields → define the field Name.
• Each field has:
- {} button: turns the field into an **object** or **array** (nested).
- Tt button: inserts **Inputs** (dynamic values the AI extracts from the conversation).
A) Simple object
1) Create a parent field (e.g., branch).
2) Click {} to convert it into an object.
3) Inside, add new values for child keys:
- branch_name = "BSAS"
- branch_id = 12
Conceptual result:
{
"branch": {
"branch_name": "BSAS",
"branch_id": 12
}
}
B) Array (list)
1) Create a field (e.g., branches) and click {} to convert it into an array.
2) Inside the array, Add new value creates items.
3) Each item can be simple or an object (click {} on the item).
Example:
{
"branches": [
{ "branch_name": "BSAS", "branch_id": 12 },
{ "branch_name": "Rosario", "branch_id": 22 }
]
}
C) Using Inputs inside the Body
• To reference an Input in the field value, use:
{input:input_name}
Examples:
Inputs (extracted by the AI from the conversation)
- name → “Person’s name”
- client_id
- city
- sku_1, sku_2
Body
- name → {input:name}
- client ({} object)
• id → {input:client_id}
• city → {input:city}
- items ({} array)
• item 1 ({} object): sku → {input:sku_1}, qty → 1
• item 2 ({} object): sku → {input:sku_2}, qty → 2
→
{
"name": "Sofía",
"client": { "id": 123, "city": "Córdoba" },
"items": [
{ "sku": "ABC", "qty": 1 },
{ "sku": "XYZ", "qty": 2 }
]
}
Body Tips
• Clearly name fields/items; avoid spaces.
• If the API expects JSON, add Content-Type: application/json in Headers.
• Use Tt to insert Inputs and avoid syntax errors.
────────────────────────────────────────
6) Inputs (dynamic parameters extracted by the AI from the chat)
• Add Input → define name and description (e.g., email, document, city).
• Reminder: within Body/Query/Headers, reference with {input:name}.
────────────────────────────────────────
7) Save and test
• Save.
• In Stage Automations, define when to use the tool.
• Test now:
- Check the Status (200/4xx/5xx).
- Review the returned Body (keys the AI will use).
- Adjust Headers/Body/Inputs if it doesn’t match the API documentation.
• In the conversation, you can see when the AI uses the tool and what the HTTP response was (status and body). This lets you easily audit and debug each call.