Mistlabs Raises $26M Pre-A as Ambi Launches AI-to-AI Communication

Mistlabs is launching Ambi beyond beta on Apple devices following a $26 million Pre-A funding round led by Luminous Ventures and Huaqin Technology. The San Francisco company is developing an ambient AI system that follows ongoing work, retains context across devices, and allows separate Ambi agents to coordinate on their users’ behalf. The financing will support development of a standalone AI wearable intended to bring that intelligence into a dedicated device.

The announcement puts Mistlabs at the intersection of three increasingly connected product ideas: personal AI memory, assistants that anticipate useful work, and systems in which multiple agents cooperate. Ambi’s proposed role is to understand a conversation as it unfolds, prepare relevant work before it ends, and carry the resulting context into the next interaction. Its agent-to-agent capability extends that ambition across people, with each agent representing the priorities of its own user.

“We’re moving beyond AI that simply waits for a prompt,” said Peter Mo, CEO of Mistlabs, in the announcement.

Funding a Software Platform and a Dedicated Wearable

The Pre-A round is backing a strategy that begins with software on devices people already use and expands into purpose-built hardware. Mistlabs identifies iPhone, Apple Watch, Mac, and MacBook as the initial platforms for Ambi’s launch, with its own wearable planned as a subsequent step.

Luminous Ventures describes its investment focus as early and growth stage startups. Co-lead investor Huaqin Technology brings a different industry background: the Shanghai company develops and manufactures smart products across mobile devices, computing, and other hardware categories. Its involvement is relevant to a startup planning a wearable, although the announcement does not identify Huaqin as Ambi’s manufacturing partner or disclose a production agreement.

Mistlabs has not disclosed a valuation, the investors’ individual contributions, or a detailed allocation of the funding beyond its stated support for wearable development.

How Context and Memory Support Ambi’s Technology

Ambi’s central product idea is that an assistant becomes more useful when it can connect the current situation to a user’s earlier conversations, preferences, and unfinished work. In the release, Mistlabs describes a system that observes relevant conversations and activities on supported platforms, develops an understanding of the user, and retains that understanding in memory. This context is intended to inform recommendations and actions as circumstances change.

The company’s website expands on that approach, presenting Ambi as a system that combines hardware, software, memory, context, and intelligence. Its emphasis is on maintaining continuity: a commitment made during a conversation should be available when a follow-up becomes useful, and an earlier insight should remain relevant when the user switches devices.

At the technology level, Ambi’s terms of service identify third-party large language models and automatic speech recognition models among the services it uses. Those disclosures establish that speech processing and language models are components of the product. They do not identify the model vendors or describe how Ambi routes individual requests between models.

That distinction helps explain where the product’s proposed value sits. A language model can interpret a request, while a speech recognition system can turn spoken content into text. Ambi’s broader challenge is connecting those capabilities to persistent context, deciding which information matters to the current task, and coordinating the work needed to produce a useful result. The release describes that behavior, but does not disclose the memory architecture, agent orchestration framework, or latency measurements behind it.

Routine: Turning a Live Discussion Into Work in Progress

Mistlabs illustrates its approach through Routine, a feature that calls on multiple agents to perform recurring work. In the company’s meeting example, a discussion about competitors triggers several activities in parallel: analyzing the conversation, checking claims, researching competing products, and organizing the findings into a draft proposal. Ambi could present that proposal on the user’s MacBook before the meeting finishes.

The practical benefit in that example is earlier preparation. A proposal normally requires someone to collect notes, identify open questions, research alternatives, and assemble a document after the conversation. Mistlabs proposes moving parts of that sequence into the meeting itself, so the participants have material to consider while the discussion is still fresh.

The company also says Ambi can continue following a project after the meeting, flagging missing information, irregularities, or new developments. Retained context is meant to let routines and ongoing work carry across a phone, watch, and computer. These remain company-described workflows; the announcement does not include results showing how accurately or consistently Ambi performs them.

AI-to-AI Communication Adds Coordination Between People

Routine involves several agents working for one user. Ambi’s agent-to-agent communication introduces another relationship: agents representing different users exchanging relevant context and coordinating within permissions those users grant.

Mistlabs describes a meeting in which each participant’s Ambi agent draws on its own understanding of that person’s preferences and priorities. The agents could identify common ground, surface potential disagreements, and suggest outcomes that account for both sides. The company characterizes this as a form of real-time mediation or negotiation while the conversation is taking place.

For example, recognizing that two people agree on a project’s goal but have different scheduling constraints could help an assistant propose a workable next step. That is an illustration of the coordination problem, rather than an additional capability demonstrated in the release. The value would depend on each agent having accurate context and sharing information its user has authorized it to disclose.

This places Ambi within the broader field of multi-agent systems, where agents divide work or coordinate decisions. The announcement’s concrete examples involve Ambi agents. It does not establish compatibility with arbitrary third-party assistants or disclose an open communication protocol.

Why Mistlabs Is Building Its Own Hardware

The wearable extends the same contextual approach into a device designed around Ambi. Mistlabs says dedicated hardware could give its system a richer understanding of the user and surrounding circumstances, making memory, routines, and recommendations more personalized over time. The product direction is therefore broader than distributing another chat interface across existing screens.

“While we’re starting with the iPhone, Apple Watch, MacBook and Mac, our move into standalone hardware sets the stage for Ambi’s intelligence to extend across entirely new device types in the future,” Mo said in the release.

The company has not disclosed the wearable’s final specifications, price, battery life, or the division of processing between the device and cloud services. Those details will determine how the hardware contributes to the experience in practice.

There is also a timing discrepancy between the announcement and the public site. The October 12 release says the wearable will follow later in 2026, while Ambi’s homepage says hardware will be available early next year, indicating 2027. Ahead of the announcement, the download page offers desktop software for Apple silicon Macs and Windows 11 x64, while listing iOS and Android as coming soon. Intel Macs are explicitly unsupported. The release describes the upcoming Apple-device launch; the current download page should not be read as confirmation that every announced mobile experience is already publicly available.

Permissions and Data Handling Are Part of the Product

A system that remembers context and coordinates work needs access to information beyond the current prompt. Ambi’s privacy notice names connectors for Google Calendar, Gmail, Google Drive, Google Docs, Google Sheets, Slack, and Lark, with access governed by the permissions users approve. It says connector data is not used to take external actions the user has not specifically requested.

Mistlabs says audio is not stored by default. Its privacy notice separately covers recordings, transcripts, memory, and other content processed when people use those features. It also says private user content is not used to train, optimize, or develop the services, while third-party AI and speech recognition providers process data to deliver requested functionality.

These disclosures matter to Ambi’s proposed shift from retaining information to acting on it. A useful assistant needs to know which context it can use, what it can share with another person’s agent, and when an action requires the user’s direction. The distinction between preparing a proposal and sending it is especially relevant as Ambi moves further into connected workflows.

With the Pre-A financing, Mistlabs has funding to pursue both its software launch and a dedicated wearable. The strongest idea in the announcement is the connection between memory and timely action: retaining the context needed to make the next step easier, then coordinating that work across devices and, potentially, across people. The launch will give users an opportunity to assess whether that continuity produces more useful assistance in everyday work.