Google is announcing today that it’s ready to let developers create “conversation actions,” little bots that you’ll be able to interact with using the Google Home speaker. Though the company wouldn’t say when Home and the Google Assistant would be fully open to third-party developers beyond “early next year,” some partners that have already been working with the company could announce new actions in the coming weeks.

The actions will let you ask the speaker a question and then have a chat about it — the example Google gave in October was ordering an Uber, then clarifying you want an Uber XL. That whole conversation actually won’t be with the Google assistant, though. Once you say the keyword for a third-party action, you’ll start talking to that company’s bot instead of Google’s. Making a decent conversational bot is hard, so Google is suggesting services like its own API.AI, which is meant to simplify the process for developers.

Google has a lot of education to do surrounding these actions — they’re a weird and new kind of audio thing that sits somewhere between a website, a chat bot, and an app. They’re similar to the “skills” on Amazon’s Alexa platform, but there are some important differences.

The first and most important difference is that Google is not going to create an “Action Store” where users can select which ones they want and “install” them on Google Home. Instead, Google itself is going to approve all the keywords that developers want to use to invoke their actions and make them all available to everybody.

That effectively means that actions will be curated by Google (like an app store), but users won’t have to install anything before using them (like the web). “It's not a direct analog to any existing ecosystem,” says Jason Douglas, director for actions on Google.

There will be app store-like policies in place to prevent things like keyword camping — one developer using another company’s name for their action or another developer trying to jump on owning important keywords like “shopping.” Douglas says that “this is a large reason for there being a review process. Much of the policy compliance is this process of name selection. “

Those policies will also prohibit “the things that you would expect policies around,” Douglas says, including “hate speech and that sort of thing.” When I asked if Google would allow something like a sex chat bot, Douglas demurred from giving a hard no, but it sounds like the available actions will hew more closely to app store policies than the the wild, open web. “The full policies will be published,” Douglas says. “It's not completely open-ended. [There are] general policies around acceptable behavior.”

Another nuance worth noting is that there are two different kinds of actions: direct and conversational. It’s only the latter that Google is inviting developers to work on. Direct actions — things that don’t require a conversation, such as turning on your lights or playing a song — still require a partnership with Google.

Put it all together and Douglas is right: this is a new kind of ecosystem, and determining the ways and rules of it is going to be complicated for Google, developers, and users. Take one example: how will users even know what actions are available? Google’s not ready to reveal that yet. The question of how Google approach “discovery and promotion,” as Douglas calls it, is also going to unanswered until “early next year.”

And so alongside Siri and Alexa, the Google Assistant lives in a weird zone that’s simultaneously more open and less open than app stores. You can get more information from more places simply by asking, but you can only ask bots that these companies have approved. And since these assistants typically only give you one answer, there’s still the fear that the answer you get will be based on partnerships instead of something more transparent — a new kind of internet bundle.

For Google’s part, that sort of closed-ecosystem is not the goal for the Google Assistant. Douglas says that “We do want to be a fair ecosystem, because that's what right for the user,” and that “we're trying to be thoughtful about it.”

But those philosophical ecosystem concerns don’t feel especially urgent at this early stage — Google just needs to get the developers’ new actions for Home working and out the door to consumers. It also needs to enable more complicated actions like buying and booking stuff. In short, it has to build an assistant ecosystem that can compete with Siri, Alexa, and Cortana.

And beyond that, it needs to ensure that talking to Home is consistent with using the other places where users can ask the Google Assistant for answers: Allo and the Pixel. Douglas says “Actions on Google will eventually be everywhere the Assistant is.”

Google says that all of it is on the roadmap and is “coming soon.” Until then, it’s focused on getting developers making things for Google Home. Amazon’s Echo may have a long head start, but Google is rushing to catch up.