Wabi recasts its mini-app platform as an invite-only agentic messenger

Why it matters

Wabi is shifting its pitch from AI-generated personal apps to software shared and used inside conversations. That puts distribution and repeat social use, rather than app generation alone, at the center of Kuyda's consumer-AI bet.

Wabi founder Eugenia Kuyda (@ekuyda) is repositioning her AI app-making platform as a messenger that creates apps and handles tasks for users and their friends. In a September 28th post on X, Wabi called the product "a new kind of messenger" and said access would begin by invitation, with users joining a waitlist through its iOS app.

https://x.com/wabi/status/2104721766690222087

The pitch puts social use and action ahead of app creation. Wabi says the software should adapt to a user's life, make apps, and get things done for groups of friends. It describes the ambition as an "OS for the agentic era," but the announcement does not specify which tasks the agents perform or how users direct them. The initial offer is a waitlist, not an open launch.

That framing builds on Kuyda's earlier thesis for Wabi: software creation should be accessible to people who do not code. She previously founded Luka, which developed the AI companion Replika. When Andreessen Horowitz announced its $20 million pre-seed investment in November 2025, the firm described Wabi as a platform where consumers could prompt, use, remix and share personal software. Its examples included fitness tools, fasting apps, clip-art generators and personal CRMs. The fundraise was led by a16z; the announcement did not disclose a valuation.

Apple's version history suggests the September announcement is a change in product framing as much as a first arrival of the underlying features. The Wabi App Store listing already described creating, customizing and sharing mini-apps. Its June 20th entry for version 2.0.4 listed agentic mini-apps, friend chat, collaborative building and remixing, and multiplayer apps. The September post calls the product Wabi 2.0, while Apple's listing records several 2.0.x releases months earlier.

The app listing also describes a platform that remains centered on mini-apps: users can generate one from a prompt, change it through conversation, share it, and use community-made tools. Apple lists the download as free with in-app purchases, including a $20 monthly Wabi Max option. The X post, by contrast, directs new users to an invite queue. These are different parts of the rollout: the app is listed, while access to the newly promoted experience is presented as limited.

For Kuyda, the social layer is central to the bet. In its investment announcement, a16z argued that Wabi could link creation and distribution, allowing users to discover and modify software made by others. A messenger could make that exchange routine: friends encounter a tool inside a conversation, use it together, then adapt it for their own group. That is a different product proposition from a prompt-based builder used alone, even if the underlying mini-app format remains the same.

The competition includes products such as Replit and Lovable, which also let users build software from prompts. Wabi's distinction, as described by its backer and its own product materials, is that it combines generation with discovery and social sharing for consumers. The messenger framing sharpens that distinction, but it also raises a practical test: whether the apps can reliably do useful things for people in a group, rather than merely give them another way to chat about software.

The September announcement makes that promise without describing a specific agent workflow or naming a task users can try. For now, the concrete product remains an iPhone app for making and sharing mini-apps, with the newly promoted messenger experience offered by invitation. Kuyda is asking users to treat personal software as something friends make and use together; the next proof point is whether Wabi can make that behavior habitual.