Wabi has transitioned its core product from a prompt-to-app builder into an invite-only AI messaging experience known as Wabi 2.0. The new platform, introduced on September 28, 2026, functions as a personal AI agent that creates task-specific, interactive interfaces on demand within a chat environment.
The pivot addresses a common limitation of standard AI chat interfaces: the “endless scroll.” In typical LLM interactions, data and tools often disappear as the conversation progresses. Wabi 2.0 instead generates durable UI components—such as calorie trackers, workout logs, and family calendars—that remain accessible and functional inside the messenger. These “on-demand” interfaces allow users to interact with specialized tools without leaving the chat or managing a library of separate standalone applications.
The startup was founded by Eugenia Kuyda, who previously created the AI companion app Replika. Wabi’s move into “personal software” focuses on providing users with utility-driven AI that can handle personal data and routines. The platform integrates with external data ecosystems, including Apple Health and digital calendars, allowing the AI to populate these custom-built interfaces with real-time health metrics or scheduling information.
Wabi’s development is backed by a $20 million pre-seed funding round closed in November 2025, led by Kleiner Perkins. This financial support has fueled the transition from a tool used to build separate apps into a unified messaging ecosystem.
Access to Wabi 2.0 is currently restricted to invited users. While earlier versions of the platform focused on its prompt-based app builder—allowing users to describe an app’s function to generate it—the 2.0 update centralizes this capability, making the AI agent the primary interface for creating and using these ephemeral or durable tools.
This evolution suggests a shift toward AI agents that possess their own user interface logic. Rather than just providing text-based responses, the system synthesizes functional widgets based on the context of the conversation. This approach aims to provide the utility of specialized software with the ease of a conversational thread.
