Wabi 2.0: The AI assistant leaves the 1-on-1 chat to organize group life

Wabi 2.0 combines group messaging, AI agents, and mini-apps generated on demand to organize everyday tasks.

Organizing a vacation among friends, splitting up family chores, or finding an outing date that works with multiple schedules: these activities typically involve a succession of messages, links, and different applications. With Wabi 2.0, Eugenia Kuyda, founder of Replika, aims to bring these exchanges and the actions resulting from them together within a single interface. Publicly accessible since October 19, 2026, following an initial invite-only release, Wabi takes the form of a messaging app in which an AI assistant can participate in conversations, perform certain tasks, and create the necessary tools for their organization. The service is available on iOS, Android, and Mac, and is currently free to use. A messaging app where the assistant becomes a participant

Most AI assistants are designed around a one-on-one relationship: a person formulates a request, and the model responds or executes an action on their behalf. This setup is suited for information retrieval or personal tasks, but it is less appropriate for situations where multiple people must decide and act together. Discussions often remain separate from calendars, reservations, or documents used to make decisions a reality.

Wabi 2.0 seeks to bring these different activities closer together. Each topic can have its own discussion thread, to which the user invites their partner, friends, family members, or colleagues. The assistant then participates in a shared space where everyone can view information and contribute to the progress of a project. A trip, for instance, can bring together destination suggestions, date selection, the itinerary, and booking steps into a single discussion.

Eugenia Kuyda presents this evolution as an answer to the limitations of exclusively individual assistants. Wabi therefore positions itself not merely as a shared chatbot, but as a system capable of moving collective activities forward. Its public launch on October 19 continues a pivot begun in late September, when the company started transforming its app creation platform into a messaging service equipped with agentic features. An app created when conversation is no longer enough

One of Wabi's key features lies in the generation of mini-apps directly inside conversations. Rather than presenting all information in message form, the assistant can offer an interface tailored to the task at hand. Preparing for a family week can thus take the form of a shared calendar, a collective decision can become a poll, and organizing travel can turn into a map accompanied by an itinerary.

This capability extends Wabi's original concept. In its manifesto, the company defends the idea of software built according to an individual's tastes, habits, and context, rather than an identical app for millions of users. The difference from the first version lies in the role taken by conversation: mini-apps are no longer just standalone creations, but tools produced when exchanges or actions undertaken require them.

The result can be saved and used within the thread, without forcing participants to manually copy information into another software program. Wabi also highlights customizable widgets for mobile home screens. This approach does not mean that the assistant systematically builds full native applications: it primarily involves lightweight interfaces generated within its own environment. Tasks that go beyond planning

Wabi claims uses that do not stop at picking a date or creating a shared spreadsheet. The assistant is presented as capable of researching options, making appointments, managing certain product returns, tracking down refunds, and completing steps with third-party services. In examples published by Eugenia Kuyda, a change in a school schedule can trigger an update to a family calendar, while an idea for a show can lead to finding a date and then purchasing tickets.

These features rely on access to the information and accounts that the user chooses to connect. Wabi notably mentions email services, calendars, contacts, photos, and certain professional services. Its privacy policy also describes potential integrations with Google Drive, Slack, Notion, GitHub, Telegram, or WhatsApp, as well as financial features and connections to health data.

The actual scope of these automations, however, depends on available services, their permissions, and the app's configuration. The presence of an integration in the documentation does not guarantee that all operations are accessible to every user or in every country. The terms of service state that certain actions, particularly purchases and bookings, may require explicit confirmation before execution. An assistant designed to anticipate forgotten tasks

Beyond requests made in a conversation, Wabi also aims to operate proactively. The company gives the example of a school form forgotten in an inbox, an altered invoice, or a request left unanswered. The goal is to enable the assistant to spot information deserving special attention, and then suggest an action or a reminder without waiting for a new instruction.

This ability depends heavily on the data to which the user grants access. The privacy policy, updated in October 2026, specifically states that Wabi can use background location when this authorization is enabled, in order to identify situations where contextual help might be relevant. Connections to accounts and applications can also enrich the assistant's preferences and memory.

The value of this approach relies on reducing repetitive organizational tasks, but it introduces a trade-off: the more information the system possesses, the more its ability to intervene depends on choices regarding privacy and permissions. Proactive assistance also requires distinguishing truly urgent information from additional notifications that could, in turn, become a source of interruptions. Personal information shared under conditions

The collective use of Wabi makes the distinction between a user's private information and information