Wabi raises $55M at a $1.25B valuation

The company presents the product as a personal agent that turns shared activities into miniapps.

The news combines funding and product positioning. It is not the launch of a new model nor a tool to generate source code.

See the official Wabi announcement.

In this article you will see:

• What Wabi announced in the round
• How shared miniapps work
• What the proposal means for those who build AI products
• How to test the app and which limits to watch

What Wabi announced in the round

In the release distributed on October 29, Wabi said it raised US$55 million in funding led by IVP, with a valuation of US$1.25 billion.

The company also cited participation from Andreessen Horowitz, Forerunner and the Amazon Alexa Fund.

The note also mentions angel investors and founders, including Tobi Lütke, Marc Benioff, Marissa Mayer, Dick Costolo and Alex Pall. According to Wabi, the new capital will be used to continue development and expand the product.

The US$1.25 billion valuation is a round data point, not a measure of revenue, user count or retention. The release does not publish independent numbers on those metrics nor comparable adoption results.

The phrase "strong interest" appears in the company's text to describe reception to the public launch. Without a user base, period and calculation method, that phrase does not allow estimating the size of demand.

The round context helps understand the bet. Wabi does not describe itself only as an individual conversation window: its proposal is for several people to share activities, miniapps and spaces around common interests.

How Wabi turns conversations into miniapps

Wabi defines the product as an AI personal assistant to use with other people. The agent combines personalized miniapps and shared spaces to organize everyday tasks, relationships and interests.

The release lists concrete examples: a group can build a weekly meal plan that becomes a shopping list, keep restaurants to try or create a trivia game for a celebration.

The company also cites travel planning, family calendars and group tasks.

The difference compared to a disposable conversation is the idea of reusing a small tool. A restaurant list, for example, can continue to exist and be updated by the group, instead of being lost in the message history.

Image credit: Wabi, official announcement.

This experience is related to other task oriented agents, but it is not the same proposal as an assistant that executes sales.

For comparison, see how the PicPay agent serves consumers on WhatsApp, an application with a commerce flow instead of shared social activities.

The release does not describe miniapps as programming projects, does not promise code export and does not present an SDK for developers.

The stated focus is to use the app to organize activities and produce small personalized tools within that context.

What Wabi's proposal means for people building AI products

The core product idea is to treat the group as part of the agent's context. Instead of optimizing only for a response to one person, Wabi wants to help multiple people plan and act around a shared activity.

The shared information can be a list, a plan, or a miniapp that the group revisits.

For software developers, this suggests evaluation points: who can edit a miniapp, how the agent resolves different versions of a plan, and what happens when someone leaves the group or asks to remove information.

These are evaluation questions, not features confirmed by the announcement. Wabi presents examples of collaboration, but does not detail permissions, change history, data export, or retention policy in the funding note.

The product also shows that "personal agent" can mean something different from an individual chatbot.

In Wabi's description, AI participates in social tasks, from building plans to turning an idea into an activity the group can revisit.

This approach is only useful when people understand what is being shared.

For agents that receive access to accounts or sensitive information, it is worth reviewing safeguards around passwords and permissions.

See the recommendations on passwords and permissions for AI assistants before connecting important data.

How to test Wabi and what is still unknown

The official announcement lists versions for iOS, Android and Mac and says a Windows version would come later.

The Wabi page centralizes the current access options; check there for availability on your platform before you start.

A simple test can use a travel plan, a meal list, or a group activity.

Watch whether the agent turns the request into a revisable tool and whether changes made by one person are clear to the others.

Test limits with non-sensitive data. Invite a few people, check what each one can see or edit, and try correcting outdated information.

Do not connect accounts, passwords, or private data before you understand the permissions and data policy.

To evaluate the product as a developer, check whether there is data export and change history, how much each miniapp can be customized, and whether the agent works without sharing the whole space.

The funding announcement does not answer these questions.

Independent adoption information is also missing. The announcement does not reveal active users, usage frequency, retention, or conversion rate of ideas into completed miniapps.

The round shows investor confidence, it does not by itself prove that the product has found recurring use.