At Build 2026, Microsoft didn’t just talk about AI agents in the abstract. It introduced Project Solara, a new operating system aimed specifically at gadgets that run agent-driven experiences—devices that don’t merely display information, but actively coordinate tasks on your behalf. The pitch is ambitious and, more importantly, very specific: Solara is “a new platform built from the ground up to power agent-driven experiences,” and it’s built on Android rather than Windows.
That last detail matters more than it might sound. Microsoft has spent years trying to make Windows feel like the center of the AI universe—through Copilot integration, developer tooling, and a steady stream of announcements about how AI will “meet you where you are.” Project Solara suggests a different strategy: if the future of computing is increasingly embodied in small, always-available devices—desks, wearables, ambient displays—then the OS layer needs to be designed around agent behavior, not retrofitted onto a general-purpose desktop-first model.
Solara is being positioned as an OS for “agent gadgets,” and Microsoft demonstrated two concept devices to make that vision tangible: a desk concept and a badge concept. Both are designed around a simple idea: agents should be reachable instantly, authenticated securely, and able to act in context—without forcing users into a traditional app workflow every time they want something done.
The desk concept: ambient, face-aware, and built for quick agent access
The desk concept looks and feels like an Amazon Echo Show–style device, but with a key twist: it unlocks with facial recognition. In other words, it’s not just a screen on a stand. It’s a device that can identify who is in front of it and then route that identity into agent workflows.
This is a subtle but meaningful shift in how we think about “interaction.” Voice assistants made ambient computing popular, but they often rely on wake words and subsequent confirmation steps. Facial recognition changes the rhythm. Instead of asking the user to initiate, the device can treat presence as a signal that the user is ready—and then bring the right agent capabilities forward immediately.
Microsoft’s framing implies that Solara isn’t only about UI polish. It’s about building an environment where agent-driven experiences can be triggered reliably and safely. If the desk can unlock for the right person, then the agent can do more than answer questions. It can start tasks that require authorization—opening the right workspace, pulling relevant context, or initiating actions that would otherwise require manual verification.
In a world where AI agents are expected to take actions (not just generate text), authentication becomes part of the interface. Solara’s desk concept makes that point visually: the first step isn’t “open an app,” it’s “be recognized.”
There’s also a design implication here. A desk device is typically shared space—home offices, conference rooms, common areas. Facial recognition is one way to reduce friction while still keeping access controlled. But it also raises the question of what happens when recognition fails, when multiple people are present, or when the user wants to switch identities quickly. Even though Microsoft presented this as a concept, the existence of the feature suggests Solara is intended to handle these edge cases as part of the platform experience, not as an afterthought bolted onto a generic Android build.
The badge concept: wearable authentication meets agent wake-up
The second concept device is a wearable badge—the kind of thing you’d use to access a work building. It includes a camera and a fingerprint scanner, and Microsoft says it can wake an AI agent.
This is where Solara’s “agent gadget” thesis becomes especially interesting. Wearables are often treated as companion devices: they notify you, track activity, or provide quick controls. But a badge is different. It’s already tied to identity and access. It’s already part of the physical security layer of many workplaces. By using that existing role, Microsoft is effectively saying: the most natural place to authenticate an agent might be the same place you authenticate yourself for doors, systems, and permissions.
A badge that can wake an agent suggests a new interaction pattern: instead of “ask the assistant,” it becomes “the assistant is ready because you are here.” The fingerprint scanner adds another layer of certainty, which is important if the agent is going to do anything beyond low-stakes tasks. If an agent can initiate actions—send messages, schedule meetings, open resources, request approvals—then the platform needs a reliable way to confirm the user’s identity at the moment of action.
The camera also hints at richer context capture. While Microsoft didn’t spell out every capability in the concept, cameras on wearables can support things like scanning documents, reading labels, recognizing objects, or capturing visual context for an agent to interpret. Combined with authentication, that means the agent can move from “conversation” to “understanding and acting” much faster.
What’s notable is that both concept devices revolve around authentication and immediate agent availability. That’s not accidental. It’s a strong signal that Solara is designed around the idea that agent experiences should be secure by default and responsive by design.
Why Android, and why that choice signals a broader strategy
Microsoft building Solara on Android is a strategic decision that aligns with the reality of device ecosystems. Android is already the foundation for a huge range of hardware categories: tablets, smart displays, wearables, embedded devices, and custom form factors. If Microsoft wants to create an OS for gadgets that run AI agents, starting from Android reduces friction for developers and manufacturers.
But there’s another angle: Android’s modularity and existing app ecosystem make it easier to integrate agent capabilities without forcing every device into a single Microsoft-centric hardware stack. Solara can potentially provide a consistent agent runtime and experience layer across diverse hardware, while still allowing familiar Android components underneath.
In practice, that could mean Solara offers a standardized way to manage agent lifecycles—how agents are launched, how they receive context, how they interact with device sensors, and how they handle permissions. It could also mean Solara defines a consistent approach to “agent surfaces,” the places where users interact with agents: screens, voice, gestures, wearables, and ambient displays.
If Microsoft had built Solara on Windows, it would have been a harder sell for the gadget market. Windows is powerful, but it’s not naturally optimized for the kinds of always-on, low-friction interactions that desk displays and badges represent. Android, by contrast, is already comfortable with constrained hardware, background services, and sensor-driven experiences.
So Solara’s Android base isn’t just a technical detail—it’s a bet that agent computing will be distributed across many device types, and that the OS layer must be adaptable enough to meet that reality.
The real innovation: designing for agent-driven workflows, not just agent chat
Most AI announcements focus on models and chat interfaces. Project Solara shifts attention toward the platform layer—the environment where agents actually operate. That’s where the hard problems live.
Agent-driven experiences require more than generating responses. They need:
1) Context management
Agents must know what’s relevant: who the user is, what device they’re interacting with, what time it is, what location they’re in, what systems they have access to, and what tasks are currently in progress.
2) Permissions and safety boundaries
If an agent can act, it must be constrained. Solara’s authentication-first device concepts suggest Microsoft is thinking about how identity ties into permissions and action authorization.
3) Reliable triggering and handoff
Users won’t always want to start an agent from scratch. They’ll want to wake it, continue a task, switch contexts, or hand off between devices. An OS designed for agents needs to make those transitions smooth.
4) A consistent “agent surface”
Whether the user interacts via a desk display or a badge, the experience should feel coherent. That implies Solara provides a framework for how agents present options, request confirmation, and show progress.
5) Background operation with user control
Agents will likely run continuously or semi-continuously to stay ready. But users need transparency and control—knowing when an agent is active, what it’s doing, and how to stop or adjust it.
Microsoft’s description—“built from the ground up”—reads like an attempt to address these issues at the OS level rather than leaving them to individual apps. That’s a big deal. When agent experiences are built ad hoc, they tend to fragment: each app invents its own permission prompts, its own context rules, its own UI patterns, and its own ways to recover from errors. A platform approach could unify those behaviors.
A unique take: Solara treats authentication as the gateway to agency
There’s a philosophical shift embedded in the concept devices. Traditional computing treats authentication as a gate to access systems. Agent computing treats authentication as a gate to agency.
In other words, once the OS knows who you are, it can decide what the agent is allowed to do—and how quickly it can do it. That’s why the desk unlocks with facial recognition and the badge uses fingerprint scanning. These aren’t just security features; they’re interaction accelerators.
This matters because agent experiences are likely to be judged by responsiveness. If every agent action requires a long sequence of confirmations, the experience will feel slow and cumbersome. If the OS can confidently authenticate the user, it can reduce friction while still maintaining safety boundaries.
Of course, this also introduces a new responsibility: the OS must handle misidentification gracefully. Facial recognition can fail; fingerprints can be unavailable; multiple people can be present. A well-designed agent OS should include fallback flows—clear prompts, quick switching, and safe defaults when identity is uncertain.
Even though Microsoft’s presentation focused on concepts, the emphasis on biometric and identity signals suggests Solara is designed to make these flows part of the core experience.
What this could mean for developers and device makers
For developers, an agent-focused OS can be a double-edged sword. On one hand, it can simplify building agent experiences by providing standardized APIs, runtimes, and permission models. On the other hand, it