Skip to content
intoCHAT

AI actions

AI actions: let your chatbot look things up in your own systems

Training covers what is written down. Actions cover what changes by the minute. Point the agent at your API and it fetches live answers, such as stock levels or delivery dates, while it chats.

Create your agent free See it answer from your own website
Example: a stock check against a shop's own inventory API

Visitor: Is the oak dining table in 180 cm available right now?

Agent: I checked our warehouse: the oak dining table in 180 cm has 3 units in stock and ships within 2 working days.

Visitor: And the 220 cm version?

Agent: The 220 cm version is out of stock at the moment. Our system shows the next delivery expected in week 44.

Visitor: Do you deliver to postcode 8004?

Agent: Yes, postcode 8004 is in our delivery area, and furniture delivery there is free.

Some answers are not on any web page

Whether a product is in stock today, what 40 units cost, or whether you deliver to a certain postcode lives in your own systems, not in a PDF. A chatbot trained only on your website has to send the visitor elsewhere. Actions let it ask your systems directly.

What actions add to your chatbot

Live data from the systems you already run, with you deciding exactly what the agent can call.

  • Live answers, not stale copies

    The agent calls your API at the moment the visitor asks, so stock, prices and availability are as current as your system.

  • Works with the API you already have

    Any HTTP method, JSON request bodies, and authentication by bearer token, basic auth, an API key in a header or query string, or custom headers.

  • The agent fills in the details

    You define the inputs, such as a product code or a postcode. The agent takes the values from the conversation.

  • You control what it sees

    Choose which fields of the response the model receives, so internal fields stay out of the conversation.

  • Actions on your own page

    Client-side actions run a JavaScript function you register on your site, in the visitor's browser, for things only your page can do.

How to set up an action

  1. Describe the action

    Give it a name and a plain description of when to use it, for example: use this when a visitor asks whether a product is in stock. The agent decides when to call it from this description.

  2. Add the request

    Enter the URL, method, headers and body, or paste a cURL command and intoCHAT fills them in, including the authentication.

  3. Define inputs and fields

    List the values the agent should fill from the conversation, such as sku or postcode, and pick which response fields the model may read.

  4. Test, then go live

    Run the action with sample values in the test runner and check the response. Then try it in the playground before visitors see it.

What you can configure

  • Any HTTP request

    Any HTTP method, with values from the conversation placed in the path, query string, headers or body.

  • Authentication

    Bearer token, basic auth, an API key in a header or query string, or your own custom headers.

  • Import from cURL

    Paste a cURL command from your API docs and the action is created with its URL, headers, body and authentication.

  • Inputs and response fields

    Typed inputs the agent fills in, and a list of the response fields the model is allowed to see.

  • Test runner

    Send a real request from the dashboard and inspect the response before the action goes live.

  • Up to 25 actions per agent

    Combine several actions, for example one for stock, one for prices and one for delivery areas.

Frequently asked questions

What is an AI action in a chatbot?

An action is a request the agent can make to another system during a conversation. You describe when to use it; the agent recognizes when a question needs it, fills in the inputs from the chat and answers with the result. In intoCHAT, actions are HTTP requests to your API or JavaScript functions on your page.

Do I need a developer to set up actions?

You need an API that returns the data, and someone who knows its URL and authentication. If your API docs show a cURL example, paste it and intoCHAT builds the action for you. Client-side actions need a few lines of JavaScript on your site.

What is the difference between API actions and client-side actions?

Custom API actions are sent from intoCHAT's servers, so they can carry secret keys safely. Client-side actions run in the visitor's browser on your page, through a function you register with window.IntoChatActions.register. Use them for things only your page can do, and never put secret keys in them.

Can actions track orders or book appointments?

There is no built-in order tracking or booking. If you run your own booking system with an API, a custom API action can call it. For order questions, the agent can point visitors to your tracking page.

Is it safe to connect my API?

Secret keys stay on the server side and are not shown to visitors. You choose which response fields the model sees, so only that data can appear in answers. Give each action the narrowest access it needs, such as a read-only key.

How many actions can one agent have?

Up to 25 actions per agent, on every plan including Free. Stores connected with Connect your store create ordinary actions, which count toward the same limit.

Connect your chatbot to your own systems

Create your agent free, add your first action from a cURL command and test it in the playground before it goes live.

Create your agent free

Free plan, no credit card needed.