Wabi, the AI agent that builds mini apps, raises $55M at a $1.25B valuation

Wabi raised $55 million at a $1.25 billion valuation, TechCrunch reported on October 29. In announcing the round, the company described itself as a "personal agent" rather than a chat app, a framing that captures how far its pitch has moved from its original product.

The company was founded by Eugenia Kuyda, who previously started the AI companion app Replika. TechCrunch reports that Wabi's focus shifted over time toward an AI messenger, and the version it now promotes, Wabi 2.0, is built around an agent that generates the interfaces you need to get things done.

Those interfaces are the "mini apps" in the headline. TechCrunch's examples include meal planning, tracking local restaurants, planning date nights, building fitness plans and designing specialized trivia games. Instead of searching an app store for a tool, the user asks the agent, which assembles something for the task at hand.

IVP led the round. "Eugenia has a unique sense for where AI is headed and a track record of great taste in building it," IVP general partner Alex Lim said in a statement, adding that she has a clear vision for how people will interact with AI, one that goes far beyond the text box or command line.

Other participants included Andreessen Horowitz, Forerunner and the Amazon Alexa Fund, along with angels including Shopify CEO Tobi Lütke, Salesforce CEO Marc Benioff, former Yahoo chief Marissa Mayer, former Twitter CEO Dick Costolo and The Chainsmokers' Alex Pall.

The round lands in a hot agent market. TechCrunch notes that the viral AI agent Instinct has pulled in another billion dollars, and that Tolan, which makes a muppet-like AI companion, is also raising. Investors are spreading bets across competing visions of what the winning general assistant looks like.

Wabi's bet is on the interaction layer. If an agent generates a bespoke interface for every task, the boundary of an app blurs: software stops being a pre-packaged set of features and becomes a shell assembled for the moment. That logic explains the pivot away from companionship and toward task execution.

Retention and cost are the open questions. Generated interfaces can fit a task closely, but they push reliability, consistency and inference spend onto the agent itself. How long $55 million funds that iteration, and whether the mini apps coalesce into reusable templates, will determine whether the valuation is borne out.