Text it like a friend: Sundae Bar's new personal AI agent takes tasks over iMessage, RCS and WhatsApp

• Who: Sundae Bar AI PLC (AIM: SBAR), an enterprise AI platform company admitted to London’s AIM market in June 2025.
• What: Sundae, a personal AI agent launched in beta on October 21, 2026, extending the company’s enterprise AI technology into the consumer market.
• How: text it on iMessage, RCS (Android) or WhatsApp — no app download, no new interface; beta details at sundae.md.
• Capabilities: its own computer, browser and persistent memory; research, content creation, digital projects and background monitoring; context retained across conversations; proactive notifications when monitored information changes.
• Demand: 150 beta sign-ups arrived “almost immediately”, prompting increased testing capacity; a waitlist has since been established.
• Origins: built on infrastructure for testing thousands of AI Skills and Agent submissions, grown out of Subnet 121 (SN121), the company’s AI development network on Bittensor — 30 development challenges, nearly 7,000 competing submissions.

The personal AI agent wars just moved into the one app everyone already opens: their text messages. Sundae Bar AI PLC (AIM: SBAR), an enterprise AI platform company listed on London’s AIM market, on Wednesday announced the beta launch of Sundae, a personal AI agent you reach by texting it on iMessage, RCS or WhatsApp — no app download, no new interface to learn.

Behind the number sits a full working environment. The company says Sundae has its own computer, its own browser and persistent memory, allowing it to conduct research, create content, build digital projects, monitor information and work on tasks in the background while you go about your day.

Demand, the company says, arrived faster than expected: 150 beta sign-ups came in almost immediately, prompting Sundae Bar to increase testing capacity, and a waitlist has since been established.

No app to install — the agent lives in your messages

Most AI assistants want you to come to them: install the app, open the tab, learn the interface. Sundae inverts that. You text a task the way you’d text a person, and it carries it out and texts you back. It retains context between conversations and proactively notifies you when something it is monitoring changes — so delegation becomes ongoing rather than a fresh prompt every time.

A computer of its own behind the number

The architecture is what separates this from a chatbot with an SMS gateway. Giving an agent its own computer and browser means it can do the work, not just answer questions: browse sources, assemble projects, keep watch on something and report back only when there is something worth saying. Persistent memory means the tenth conversation starts where the ninth left off.

Born from an agent-testing rig, not a product lab

The consumer product is a byproduct of an enterprise problem. To evaluate thousands of AI Skills and Agent submissions for its commercial platform, Sundae Bar built an automated environment where agents operate computers, use browsers and tools, complete tasks and have their performance assessed. That infrastructure grew out of Subnet 121 (SN121), the company’s AI development network on the Bittensor network, which lets developers around the world compete to build better AI capabilities — 30 development challenges and nearly 7,000 competing submissions to date, according to the company.

CEO Jill Kenney framed it as leverage: “We’ve spent the past year building technology that allows AI to do the work, not just answer questions. With sundae_bar, we’re bringing that capability to businesses. Now, with Sundae, we’re bringing it to individuals.” Her punchline: “One technology foundation, two distinct markets.”

Why it matters: agents are escaping the chat tab

This is part of a broader pattern. Agent modes on phones, voice-agent stacks, payment rails for agent purchases — the frontier has moved from what the model can say to where the agent can act. A personal agent with its own browser sitting inside iMessage is a bid for the most contested real estate in computing: the messaging habit. The open question is the one haunting every agent product — what happens when a delegated task goes wrong in a medium where there is no visible screen to watch it work.

For now the company is treating Sundae as an experiment. It remains in beta, the company says, focused on testing functionality, understanding user behaviour and assessing commercial opportunities — while the core enterprise business, a subscription platform plus a marketplace of more than 700 AI Skills and Agents and bespoke deployments, carries on unchanged.