Microsoft unveiled Project Solara at its Build 2026 conference as a strategic effort to move computing from app-driven endpoints toward devices built around cloud-backed AI agents. The announcement framed Solara as a chip-to-cloud platform and a set of reference device concepts intended to host persistent, task-oriented agents that act on users’ behalf rather than surfacing traditional apps.
The company described Project Solara as centered on a lightweight, enterprise-grade device platform called the Microsoft Device Ecosystem Platform (MDEP), built on Android Open Source Project (AOSP) foundations, plus an Agent Shell and cloud services for agent lifecycle and orchestration. Microsoft positioned the initiative as a joint effort with silicon partners to produce a turnkey path for manufacturers to ship “agent-first” devices.
Project Solara: a new foundation for agent-first devices
Project Solara rethinks the device as a host for agents, software entities that maintain context, execute workflows, and interact with cloud services, instead of as a platform for user-launched apps. That shift reframes the primary UI as the agent interaction model and surfaces outcomes rather than application windows.
Under the hood Microsoft is delivering MDEP (the Microsoft Device Ecosystem Platform), which the company describes as a secure, minimal runtime layer atop AOSP that hosts an Agent Shell capable of dynamically loading and composing cloud agents. The Agent Shell and an adaptive access layer are designed to let agents interface with services, sensors and enterprise systems in a controlled way.
Microsoft emphasized that Solara is not a consumer-only experiment: early demonstrations and planned pilots target enterprise scenarios and vertical integrations across healthcare, retail and logistics. The company named multiple pilot partners and hardware collaborators as part of the initial rollout conversations.
Technical architecture: chip-to-cloud and the agent shell
The technical core of Solara is the “chip-to-cloud” model Microsoft presented at Build: coordinated work across silicon, a slim device OS, and cloud-native agent orchestration so that agents can run persistently and efficiently at the edge while leveraging cloud compute when needed. This co-design approach aims to optimize latency, power and privacy trade-offs for always-on agent experiences.
MDEP is explicitly distinct from Windows. Microsoft framed it as a purpose-built runtime for constrained, specialized devices, not a replacement for Windows on PCs, giving device makers an AOSP-based, enterprise-orientated option to host agent-centric experiences. The platform includes hooks for identity, attestation, and secure communication with Azure-hosted services.
On the developer side Microsoft positioned existing tooling, Copilot Studio, Microsoft 365 Agents SDKs, and Azure orchestration services, as the path for building Solara-ready agents, so organizations that have adopted Copilot and Microsoft cloud tooling can reuse investments while targeting new device form factors.
Reference devices and pilots: badges, desk hubs and new form factors
At Build Microsoft showed concept devices to illustrate what agent-first hardware might look like: a palm-sized smart badge, a small card-like wearable with camera and biometric sign-in, and desk-oriented hubs that act as ambient assistant surfaces. Those reference designs emphasize dedicated sensors, low-power connectivity and always-available agent interfaces rather than full smartphone feature sets.
The badge concept demonstrated contextual workflows: a clinician scanning into a chart and invoking an agent to fetch patient context, a warehouse worker snapping a package image and having an agent update inventory and shipping records. Hardware details shown in early demos included Qualcomm-based SoCs, cameras, fingerprint sensors and 5G connectivity in some concepts.
Microsoft said it would move from concept to private pilots with industry partners, positioning Solara as a platform for enterprise deployments where a constrained, secure device running tailored agents can replace multi-step app flows. The partner list and pilot plans were highlighted as evidence the platform targets real-world, vertical use rather than solely consumer experimentation.
What Solara means for apps, Windows and the developer model
Analysts and commentators framed Solara as part of a broader “agents replace apps” narrative: if agents can perform tasks proactively and compose services, the traditional app model, where users open an app and perform actions, could shrink in importance on certain device classes. Observers called this an inflection point in human-computer interaction design.
Microsoft’s messaging made clear that Windows and its investments are not being abandoned: Windows will continue to evolve with AI features, while Solara targets a different set of trade-offs and device types. Enterprises should therefore expect a heterogeneous future in which Windows, cloud Copilots and agent-first devices coexist and integrate.
For developers the shift implies new priorities: designing stateful agents, managing identity and consent across distributed surfaces, and optimizing agents for intermittent connectivity and constrained hardware rather than building one-size-fits-all apps. That will require updated SDKs, testing frameworks and new UX patterns for agent-grounded interactions.
Enterprise use cases: healthcare, retail, logistics and frontline work
Microsoft highlighted vertical scenarios where agent-first devices could remove friction from routine tasks, for example, healthcare clinicians using a badge or hub to surface patient context and document encounters, or retail staff using a wearable to handle inventory and checkout workflows without switching between apps. The company pointed to its Dragon Copilot and integrations as early examples of agent-led clinical workflows.
In logistics and retail the promise is similarly operational: quick item scans that trigger agent actions (update records, route packages, schedule pickups) and contextual customer assistance on the floor without multi-step mobile app interactions. The design emphasis is on minimizing distraction and task-switching for frontline workers.
From a productivity standpoint, agents on dedicated devices can also automate cross-system orchestration, for example, summarizing actions, queuing follow-up tasks, and handing off complex operations to cloud services, which could materially reduce administrative over in enterprise workflows.
Security, privacy and governance considerations
Microsoft framed MDEP as an enterprise-grade platform with built-in device attestation, secure identity and managed cloud endpoints to mitigate many of the standard edge risks. The company emphasized creating controlled interfaces that let organizations govern which agents can access systems and data.
That messaging sits alongside broader concerns raised by observers: always-on agents raise questions about surveillance, data minimization, and user consent, and Microsoft has publicly addressed worries about attention economics and potential over-reliance on AI. Microsoft leadership reiterated commitments to responsible deployment even as it pushes agent-first experiences.
Regulators, enterprise security teams and standards bodies will need to define expectations for provenance, explainability and auditability of agent decisions, especially in regulated domains such as healthcare and finance. Early pilots will be as much about proving safe governance as demonstrating technical value.
Adoption barriers and the path a
Successful adoption depends on coordinated work across silicon partners, OEMs, cloud providers and enterprise software teams. Qualcomm and other silicon partners were presented as central collaborators in the Solara vision, underscoring that hardware integration and power/latency optimization are core to the platform’s promise.
Beyond hardware, the ecosystem question looms large: enterprises must invest in agent design, identity integration, and operational controls, while developers weigh whether to adapt existing services to agent models or wait for more mature tooling and standards. Fragmentation remains a real risk if multiple, incompatible agent platforms emerge.
Microsoft indicated private pilots with retailers and healthcare partners in 2026 to validate scenarios and iterate on the platform. Those pilots, and the data they produce about usability, security and ROI, will determine whether Solara becomes a niche set of specialized devices or a broader foundation for a new class of ubiquitous, agent-driven endpoints.
Project Solara signals Microsoft’s bet that the next wave of compute will center on agents that act for users across contexts, rather than single-purpose apps on general-purpose devices. For enterprises and policymakers the immediate priority is to engage with pilots, evaluate governance models, and measure where agent-first devices actually reduce cost and friction.
If Solara’s pilots validate the model, expect to see a steady stream of specialized hardware and tailored agents in regulated verticals first, with broader consumer or worker-oriented deployments following once security, privacy and developer ecosystems mature. The coming 12,24 months of pilots and standards work will shape whether agent-first devices become a foundational platform or a series of isolated experiments.
