Unlike traditional AI assistants that primarily respond to requests, Wabi is intended to identify useful tasks and take initiative. With permission to access connected services, it can organize information from emails and calendars, suggest activities, track ongoing responsibilities, and generate small applications without requiring users to build them manually.
Kuyda told TechCrunch that the company's approach reflects lessons from Wabi's original product, which focused on letting people create and share applications using natural-language prompts. The team found that users valued personalized software but were less interested in spending time creating it themselves.
Wabi 2.0 shifts more of that work to the assistant, which can determine what tools would be useful based on a user's needs and connected information. Linking an email account, for example, can prompt Wabi to establish relevant conversations and applications automatically.
The interface allows each conversation to contain its own collection of tools, such as calendars, trackers, maps, recommendations, and planning applications. Those tools remain accessible within the relevant thread instead of being buried among previous chatbot messages.
Kuyda has demonstrated the system using conversations from her own life. Her health thread includes tools for monitoring exercise, weight, nutrition, and medical appointments. A family conversation brings together her daughter's schedule, school information, nearby activities, and a Japanese-language learning app.
Another shared conversation with her boyfriend contains restaurant recommendations, entertainment suggestions, a relationship question delivered each day, and information from their shared calendar.
The company also highlights practical uses such as organizing family schedules from school emails, preparing travel itineraries, responding to weather-related changes, and coordinating bookings. Wabi can assist with appointment scheduling, returns, forms, subscription monitoring, and identifying unexpected bill increases or outstanding credits.
A central feature is the ability to include multiple people in conversations with the assistant. Rather than requiring each person to manage separate AI interactions, Wabi allows participants to coordinate tasks and access shared tools in a common thread.
Kuyda argues that this structure addresses a limitation of assistants built around one continuous conversation, where accumulated messages can make information difficult to retrieve.
“Single thread experiences, if you think about it, get worse over time because it’s just so much context in one endless conversation,” she told TechCrunch.
Wabi also allows users to control what personal information they share within individual conversations. Available connections include email, calendars, contacts, location, photos, and health information, with permissions managed by the user.
According to Kuyda, access to personal data is limited to the relevant conversation, and participants can see when another person requests information through the assistant. Wabi also seeks approval before taking certain consequential actions, while connected services can be disconnected.
The Android version includes optional location-based assistance, read-only access to Health Connect information, and the ability to initiate voice or video calls from conversations. Wabi also supports home-screen widgets on mobile devices and use across computers and phones.
The company's handling of personal information is particularly relevant because its assistant can work with sensitive details shared across multiple people. Kuyda said Wabi operates through cloud infrastructure, stores most user data in the United States, and has completed SOC 2 compliance requirements involving an independent assessment of its security controls.
Wabi's public release follows several weeks of expanded availability. The company initially introduced the redesigned product with a waitlist on September 28, followed by Android availability on October 6 and a Mac release on October 14. Users can now download the app without requesting an invitation.
The startup's development builds on Kuyda's earlier work in conversational AI. Before creating Replika, she founded Luka, which developed a chatbot concierge called IO for tasks such as restaurant recommendations and travel planning. Kuyda presented the concept at TechCrunch Disrupt in 2014, years before launching Wabi.
Replika subsequently became known for AI companions designed around personal conversations. Wabi takes a different approach by emphasizing task completion, personalized applications, and interactions involving multiple people.
The company raised $20 million in pre-seed funding led by Andreessen Horowitz, announced in November 2025. At that stage, Wabi was positioned primarily as a platform for creating and sharing AI-generated applications. Its latest version retains those capabilities while placing them inside a messaging experience.
The assistant enters a market that includes personal AI products such as Meta's Muse and Instinct, which also aim to handle tasks across connected services. Kuyda sees Wabi's combination of collaborative conversations and generated applications as its main distinction.
“It’s doing what Instinct and Muse do, but it also has all these apps, which helps users do more stuff,” she said.
Wabi is currently free, although the company has not finalized how it intends to generate revenue. Kuyda also told TechCrunch that the startup is discussing additional fundraising.
For now, Wabi's public launch gives users access to an assistant that combines messaging with software created around their activities. Its defining feature is the ability to organize those capabilities across shared conversations, allowing the AI to help groups manage tasks and make plans together.
This analysis is based on reporting from TechCrunch & Unite.ai.
Image courtesy of Wabi.
This article was generated with AI assistance and reviewed for accuracy and quality.
