Agent Actions

Who can use this?

• Owners and Admins. Custom roles with Can build and deploy Agents or Can manage Agents.
• Configure actions on the agent’s Actions tab, then mention them in Role, Goal, or Instructions with @.

​

Plan Availability

​

Two kinds of actions

• System actions. Built in. They already sit under System Actions on the Actions tab: assignTo, getCurrentDateTime, and checkInstagramUserFollow.
• My Actions (custom). You build these yourself under My Actions by clicking Create Action: API Call and Code Execution.

​

assignTo

1. On Home, type @assignTo in Instructions.
2. A picker opens right there. Select the user, team, or bot.
3. Gallabox inserts a tag for that target (for example, @assign_to_sales_team).

“If the customer asks about billing, refunds, or invoices, use @assign_to_billing_team.”

​

getCurrentDateTime

“Use @getCurrentDateTime when customers ask if we are open right now.”

​

checkInstagramUserFollow

“If the customer says they followed our page, use @checkInstagramUserFollow before giving the 10% discount code.”

​

API Call

• Look up information and reply with it (check an order status, check stock, look up an account).
• Send information out to another tool (save a new lead in your CRM).

​

Step 1: Create the API Call action

1. Open your agent and go to the Actions tab.
2. Under My Actions, click Create Action.
3. Fill in the top bar:
   • Action name: give it a clear name, such as check_order_status. Spaces turn into underscores automatically.
   • Action Type: select API Call.
   • Process in background:
     • Turn this OFF if the agent needs the API answer before replying to the customer (for example, checking an order status).
     • Leave it ON only when the agent is sending data out and does not need to wait for an answer (for example, logging a lead).
   • Action description: write in plain English what this action does and when the AI should use it. Example: Fetches order status and delivery date using the customer’s order number.

​

Step 2: Add your API link and details

1. Pick the Method Type (GET to fetch data, POST to send data, PUT, PATCH, or DELETE).
2. Enter the Request URL.

https://api.yourstore.com/orders/{{order_id}}

• Type: String (text), Number, or Boolean (true/false).
• Description: tell the AI what to ask the customer for. Example: The order number shared by the customer.

​

Step 3: Set authentication and body (if needed)

• Headers tab:
  • If you already connected the app in Gallabox (such as HubSpot, Zoho, Shopify, or Razorpay), pick it under Authentication and choose your saved connection under Select connection (or click Add New Connection).
  • For any other tool, leave Authentication as None and add your API key under Header (for example, Key Authorization and Value Bearer your_api_key).
• Body tab (shown for POST, PUT, PATCH, and DELETE):
  • Choose json, www-form-urlencoded, or form-data.
  • You can put {{variables}} inside the body too, and describe them under Variable Configuration.

​

Step 4: Mention it in Instructions

“When a customer asks where their order is, ask for their order number, then run @check_order_status and share the status and delivery date.”

​

Code Execution

​

How to set up Code Execution

1. Go to Actions → Create Action.
2. Set Action name (for example, calculate_shipping) and pick Action Type → Code Execution.
3. Turn OFF Process in background if the agent needs the result to reply to the customer.
4. Write a clear Action description (for example, Calculates shipping cost from weight in kg).
5. Under Function Parameters (Input Schema), click Add Parameter for each value the code needs:
   • Field Name: the name used in the code (for example, weight).
   • Data Type: String, Number, or Boolean.
   • Value Source: pick AI Provided if the agent should get this from the customer’s chat. Pick Static value if it never changes (like a fixed tax rate).
   • Description: explain what the AI should collect (for example, Package weight in kg).
6. In Code Editor (JavaScript), write your logic inside function execute(inputs) and return an object:
   function execute(inputs) {
     const cost = inputs.weight * 50;
     return {
       shipping_cost: cost
     };
   }
   
7. Under Output Schema, click Add Property and list every field your code returns (in the example above, Property Name shipping_cost, Data Type Number). The AI only reads fields listed in Output Schema.
8. Click ► Test Run, enter a sample number, and click Run to check the output before saving. If you see an Extra fields detected warning, add those missing field names to Output Schema.
9. Click Save, then mention @calculate_shipping in Home → Instructions.

​

FAQs