Actions | Ordemio Help Center

What are Actions?

Actions let your AI agent interact with external systems through APIs. They allow the agent to fetch real-time data or perform operations such as refunds, cancellations, or customer lookups.

Without Actions, the agent can only answer based on its training or your Knowledge Base. With Actions, it can actually work with your live systems.

⚠️ Note: Adding Actions does not fine-tune the AI. They are tools the AI can call when needed.

How to set up an Action

To add a new Action, go to Ordemio Dashboard → Copilot → Actions and click Add Action.

Fill out the form step by step:

1. Active

What it is: The Action activity toggle.
How to use it: Switch this on when the Action is ready for production use. During testing, you can keep it disabled — the AI agent will not use this Action while it is inactive. This makes it easy to refine settings without removing the Action completely.

2. Name

What it is: A unique name for your Action.
How to fill it out: Choose a clear, descriptive English name such as get subscription or create user info.

Important: Action names are automatically converted to kebab-case.

Kebab-case is a format where all letters are lowercase and words are separated by hyphens (-), like meat on a skewer.

Example:

Entered name: Get User Subscription
Used by the agent: get-user-subscription

3. Description

What it is: A short explanation of the Action for the AI agent.
How to fill it out: Write in simple language so the agent understands:

What the Action does.
When it should be used.
What information it returns or what it performs.
Any extra instructions (e.g., when not to use it).

Important: The agent relies heavily on this description to decide whether the Action is relevant. The clearer you are, the better the results.

4. Request

What it is: The technical details of the API call.
How to fill it out:

Method → select the correct HTTP method (GET, POST, PUT, PATCH, DELETE) based on your API docs.
Endpoint URL → paste the full URL where the request should be sent.

Example (from Stripe):

Method: GET Endpoint: https://api.stripe.com/v1/customers/search

5. Parameters

What it is: The list of inputs the agent must provide for the API request.
How to fill it out: Define parameters in JSON Schema format, specifying:

type (e.g., string, number)
description (clear explanation of what it is and where the agent finds it)
enum (allowed values, if applicable)

Example:

{ 
 "query": { 
  "type": "string", 
  "description": "Customer search query by Stripe format. Example: email:'[email protected]'" 
 } 
}

6. Headers

What it is: The HTTP headers required by your API.
How to fill it out: Specify all necessary headers in JSON format. Most often, this includes:

Authorization → e.g., Authorization: Bearer YOUR_API_KEY
Content-Type → usually application/json

Example:

{ 
"Content-Type": "application/json", 
"Authorization": "Bearer XXX-XXX-XXX" 
}

How the agent uses Actions

The agent reviews the descriptions of all Actions to decide when one is relevant. You can also explicitly call Actions inside Workflows by writing Call tool: action-name. End users never see the Action call — only the final, natural reply.

Use cases

Look up a customer in Stripe.
Retrieve subscription details.
Cancel or pause a subscription.
Trigger a refund by charge ID.
Pull CRM or ERP data into a chat.
Fetch invoices, orders, or shipping status.
Create a support ticket in Jira or Zendesk.

💡 Best practice

Keep Action names simple and consistent so they are easy to reference in Workflows. Write descriptions in natural language, since that text is how the AI decides when to use an Action. Test Actions while inactive and only enable them once verified. Always document parameters clearly and use authentication headers for security.

💡 Business value

Actions let your AI deliver accurate answers by pulling live data directly from your systems. They reduce manual effort, automate routine lookups, and make the agent a reliable extension of your operations.

Why use Actions?

Actions expand what your AI can do. They turn your agent from a responder into an operator — able to fetch information, trigger events, and interact seamlessly with the tools your business already uses.