The Google Assistant is getting an API today with the launch of "Actions on Google," a way for developers to build "conversation actions" that can be called up on the Assistant's various interfaces.
The "Actions on Google" API was announced during the Google I/O 2016 keynote and has since existed in some form as a private early access program. Today Google is throwing the doors open to developers, who can create Google Assistant commands using the chatbot developer tools API.AI and GupShup.
The "Google Assistant" is available as a voice-command system on the Google Pixel and Google Home and as a chatbot in Google Allo, Google's new instant messaging app. While the branding "Assistant" suggests they are all the same system, we found out in our various reviews that they definitely are not. They are all slightly different implementations of the same idea, with some commands working in some interfaces and not others. Sure enough this "Actions on Google" API is only launching on one of the interfaces: Google Home. Google's blog post says it will "continue to add more platform capabilities over time, including the ability to make your integrations available across the various Assistant surfaces like Pixel phones and Google Allo."
How it works
Users won't be able to actually "install" these new actions on their devices. Apparently the third-party commands are enabled centrally in the cloud, by Google, which are then accessible on all Assistant devices. You can see how this works with the two new developer demo commands, “talk to Number Genie”—a number guessing game—and “talk to Eliza"—an old chatbot experiment from the '60s. These new commands will magically, silently work on your Google Home starting right now. (We're still not sure how users are supposed to discover and remember these new capabilities.)