AI App Maker Wabi Pivots Into Personal Agent Messaging Platform

Wabi is shifting from its original AI app-building model toward a messaging-based personal agent experience that can still create apps and interfaces on demand.

The startup’s new direction, called Wabi 2.0, is built on the idea that users should be able to talk with an AI agent while also using software-like tools generated for specific tasks.

Wabi Moves From App Building to Personal Agents

Founder Eugenia Kuyda, who previously founded AI companion company Replika, described the new version as a personal agent that can complete tasks and create the interface a user needs in the moment.

The shift moves Wabi closer to the growing personal AI agent market rather than positioning it mainly alongside prompt-based app builders. The company wants users to combine conversations, productivity tasks, and interactive software within the same experience.

In Wabi’s announcement, the company outlined the new messaging-focused experience and its broader vision for AI-generated interfaces.

Interfaces Can Be Created Around Individual Tasks

Kuyda argues that chat-only AI experiences become difficult to manage as conversations, tasks, and history accumulate in a single stream. Wabi’s approach is to let the agent create visual interfaces that users can tap, scroll through, and revisit.

Examples could include a calorie tracker, workout log, family calendar, language-learning tool, or an app for following local events. These tools could appear within relevant conversations rather than existing as separate standalone apps.

In Kuyda’s explanation of Wabi 2.0, she said the company sees the future as combining a user’s agent, apps, and personal connections in one place.

Wabi 2.0 Starts With Invite-Only Access

The new Wabi 2.0 experience is currently available by invitation, with access codes being distributed through X.

The pivot reflects Wabi’s attempt to combine the flexibility of AI agents with the structured interfaces people already expect from traditional software.