Use the AI section of Atoms Cloud to choose an AI model, use it in your app, and review the app’s AI usage.

This page explains AI features that users interact with inside your app. Atoms Cloud provides the model library and project usage information, while the Atoms Agent in Chat builds and updates those features.

Understand AI in Atoms Cloud

Open the project in the Atoms desktop version and make sure it is in Web mode. The AI Library contains the models currently available for the project, while AI Usage shows project credit usage and active models.

If Atoms Cloud is not connected or AI is unavailable, ask the Agent to connect Atoms Cloud or follow Atoms Cloud before continuing.

Open the AI Library in Atoms Cloud

1. In the same workspace toolbar where the App Viewer is, select Atoms Cloud.
2. In the sidebar, select AI.
3. Open AI Library.

The AI Library shows the models currently available for your project. You can browse the category filters or use Search to find a suitable model.

The library can change over time. Use the models and information shown there rather than relying on an older model or provider list.

Add a model to your request

1. Find a model in AI Library.
2. Select # Add on its card. Atoms inserts the model tag into the current Chat message.
3. Complete the feature request after the tag, then send the message to the Atoms Agent.

Select # Add to tell the Agent which AI model you want to use. The model appears as a tag in your Chat message. You can then describe what you want the Agent to build or change.

You can also attach a model directly from Chat: type #, select AI, then choose one of the currently available models.

Select # Add to insert a model tag into your Chat request.

Describe the feature clearly

After attaching a model, tell the Agent how you want the feature to work from the user’s perspective. Include the following details:

What to describe	What to tell the Agent
Purpose	Explain what the feature should help users accomplish.
User journey	Describe what users will enter, upload, or select, and where they should see the result.
Information used	Specify what information the AI should use and anything it must not include.
Expected result	Describe the preferred format, tone, language, length, and any important limits.
Loading and errors	Explain what users should see while waiting, when information is missing or invalid, and if a request fails.

You do not need to explain how the feature should be coded. Focus on what users do, what result they should receive, and how the app should respond at each stage.

The generated layout may differ between projects. Check the behavior you requested rather than expecting every app to use the same design.

The selected AI model appears as a tag in the Chat composer.

Test the feature in the App Viewer

After the Agent finishes, test the complete user journey in the App Viewer. Use harmless sample content that does not contain personal or sensitive information.

Test	What to do	Expected result
Normal input	Submit a typical example that the feature is designed to handle.	The requested output appears in the correct place and format.
Empty input	Try to continue without providing the required content.	The app explains what the user needs to provide and does not send an unnecessary request.
Invalid input	Provide content in an unsupported format or outside the feature’s expected requirements.	The app remains usable and displays a clear explanation of the problem.
Loading state	Submit a request and observe the app while it is being processed.	The app shows that the request is in progress and prevents accidental duplicate submissions.
Failed response	If a request fails during normal testing, check how the app responds.	The app displays a useful message, allows the user to try again, and does not expose technical details or private information.

Do not deliberately use sensitive data, consume all available credits, or disrupt the service to create an error. If you cannot safely reproduce a slow or failed response, tell the Agent how the app should handle that situation instead.

Review AI usage in Atoms Cloud

1. Return to Atoms Cloud → AI.
2. Open AI Usage.
3. Under Overview, review Total Credits Used and Active Models for this project.
4. Review the models listed under Active Models.

AI Usage shows a project-level overview. For wallet balance, top-up, and wider Cloud or AI usage details, see Cloud & AI Wallet.

To review the credits used by a specific Agent response, select ... below that response, then review Credits used. This message-level value is separate from the project overview in AI Usage.

Open the message menu to review the credits used by that Agent response.

Screenshot to add: AI Usage in a clean test project, showing Overview, Total Credits Used, and Active Models. Highlight the AI Usage tab in orange. Keep a current model name visible so the list is understandable, but mask account details, workspace information, and any balance information that is not required. Caption: “Open AI Usage to review project credit usage and active models.” Alt text: “AI Usage tab showing project credits used and active models.”

Protect user data and credentials

Content submitted through an AI feature may be sent to the selected model provider for processing. Only send the information needed to complete the request.

Never send passwords, API keys, access tokens, or unnecessary personal or confidential information. Make sure users understand when their content is being sent to an AI model, and follow your organization’s privacy and data-handling requirements.

Always review AI-generated results before relying on them for important decisions. See the Atoms Terms of Service for more information.

If your app needs a private credential from an external provider, store it using Keys and secrets. Never paste the credential into Chat or include it in code that runs in the browser.