Add Custom Capabilities with the Smart Home AI Toolkit
To support the unique features of your device, you can build custom smart home interfaces. You can combine these custom capabilities with pre-built Alexa interfaces to create a virtual device that matches the features of the product you want to connect to Alexa.
The Smart Home AI Toolkit helps you integrate your smart home devices with Alexa+ using AI-guided development. You can describe your device conversationally by chatting with the Alexa+ smart home agent to create a new capability from scratch. Or, you can start with an existing specification and chat with the agent to fill in the gaps. The agent generates the custom capability, including the interface definition (directives, properties, and responses), test utterances, and UI for the Alexa app and Alexa-enabled devices with a screen.
After you publish the custom capability, you can add it to a new or existing smart home add-on, following the same process you use for pre-built Alexa interfaces. Alexa+ infers the correct capability interface from customer utterances.
At this time, you can create custom capabilities in English (US) only.
Choose your starting point
The console provides two starting points. Both lead to the same review, testing, and submission steps.
- Chat and Build – A guided, conversational flow with the Alexa+ smart home agent. Choose this option when you want to describe your device in plain language and let the agent ask clarifying questions. This method doesn't require technical schema knowledge and is the recommended starting point.
- Upload Specification – A document-driven flow. Choose this option when you have formal technical documentation, such as requirements, API reference, or JSON manifest file. Use this method if you want the agent to extract as much information as possible automatically, and then ask only about gaps.
Model capabilities
Using the Alexa+ smart home agent, you can build individual capabilities and create a device that includes the appropriate custom and pre-built capability interfaces that your product supports. The agent generates sample voice utterances and visual experience for you.
For an example pet feeder, you might build a custom capability using the Smart Home AI Toolkit to fill the food and water bowls and also use the pre-built Alexa.PowerController capability to turn on the feeder. You combine these capabilities into your pet feeder device and publish the capability to the Alexa registry. After publication, Alexa+ can associate the capabilities with your device.
Best practices
Follow these best practices to create and publish a custom capability interface.
- Describe your device capabilities conversationally in English. The agent interprets natural language more effectively than cryptic names or shorthand.
- Alternatively, use detailed documentation about your device, feature, or scenario and chat with the agent in English to fill in the gaps.
- Use high-quality descriptions for your actions. These descriptions help the agent find the right interface for the utterance.
- Make sure that the properties support customers in different locations. For example, for temperature, best practice is to support both Fahrenheit and Celsius.
- The accuracy of the generated custom capability depends on, among other things, how well the sample utterances reflect real-world language. Verify that the generated utterances use natural language to help Alexa identify the correct directive to invoke.
- If your capability defines reportable properties, your smart home add-on must report state changes. For more details, see Understand State and Change Reporting.
Related topics
Last updated: Jul 22, 2026
