How DroiClaw is building an AI-native operating system for the agentic era

Artificial intelligence, once a cool feature living exclusively inside your favorite app, is upgrading. Now, the shift is toward using AI as the foundation and not simply an added feature.

AI is now becoming part of operating systems and amping up how we use and interact with devices. The vision: instead of requiring people to open apps and complete tasks manually, cutting-edge next-gen AI platforms aim to understand intent and act on a user’s behalf with minimal intervention.

This hands-off approach reflects one of the core visions behind DroiClaw (Zhuge). Developed by Shanghai Droi Technology, DroiClaw aims to build an app-free Agentic OS ecosystem supported by a hybrid on-device and cloud AI architecture, with an emphasis on creating a secure, controllable, and observable environment for intelligent agents.

What makes an operating system AI-native?

It’s not uncommon to see mobile platforms offering AI-powered image editing, writing assistants, and voice features these days. However, in most cases, these capabilities are layered onto a traditional operating system that still revolves around individual apps.

DroiClaw’s approach is fundamentally different. There’s AI embedded right into the operating system’s underlying architecture. What does that mean for the end user? AI functions are no longer isolated to apps, and instead the platform integrates AI into system scheduling, user interaction, task execution, and resource management – it’s holistic. This enables AI to become part of how the operating system itself works, rather than another feature users have to launch.

DroiClaw’s AI-centric architecture naturally translates to a next-generation Smart terminal OS. With simultaneous support for multimodal interactions through voice, text, images, files, and video, and the ability to adapt to users’ habits over time, the technology is intriguing.

From apps to an agentic OS

For decades, smartphones have been organized around applications, but these apps take work. Whether booking travel or summarizing documents, the workflow often involves opening multiple apps and moving information between them.

DroiClaw’s vision is to gradually move beyond this model by creating an app-free Agentic OS ecosystem. Rather than expecting users to navigate individual applications, AI agents are designed to:

Understand natural-language requests

Determine the steps needed to complete them

Coordinate system functions and compatible services

Carry out approved actions automatically

The emphasis is on reducing unnecessary manual tasks without giving up control completely. Sensitive operations continue to require authorization, but routine workflows can become significantly more streamlined as agents coordinate tasks across the operating system.

Agentic OS platforms diversify the operating system’s focus. Computing becomes increasingly intent-driven. A traditional OS will ask which app you want to use, whereas an AI-Native OS first seeks to understand what it is you want to accomplish.

Hybrid on-device and cloud AI

Performance, privacy, connectivity, and computing power all play an important role in delivering responsive AI experiences. DroiClaw uses a hybrid architecture that combines on-device AI with cloud-based large language models to support different usage scenarios.

Under normal conditions, users primarily rely on cloud-based large models, which provide more advanced capabilities, including:

Multimodal interactions

AIGC and content generation

Complex information processing

Skill execution and other advanced AI functions

In specific situations, such as when network access is unavailable or cloud service credits have been exhausted, the system can automatically switch to an on-device model to process basic queries. The on-device model primarily supports basic question-and-answer functions .

This hybrid setup allows DroiClaw to maintain access to essential AI functions when cloud services are temporarily unavailable, while cloud-based models remain the primary option for more advanced use cases.

The platform also supports multiple cloud-based AI models, including customized and privately deployed options, giving users greater flexibility in choosing the AI services that best suit their needs.

A secure, controllable, and observable agent ecosystem

Transparency and control are must-haves as AI agents become capable of performing increasingly sophisticated actions.

DroiClaw aims to foster an AI Agent ecosystem where automated actions remain transparent, manageable, and secure. User authorizations protect any sensitive operations, while permission management, task scheduling, and layered data protection set clearly defined AI agent boundaries.

The platform also looks to emphasize operational visibility so users and collaborators can better understand the task execution framework. Combine that with local processing for appropriate privacy-sensitive scenarios, and these safeguards are intended to help build trust as AI agents become more deeply integrated into everyday computing.

Affordable AI and an open platform

Shanghai Droi Technology also values AI democratization – the goal is to make advanced AI capabilities available beyond premium flagship devices.

The first smartphones shipping with DroiClaw, which currently include devices from Coolpad and Philips, are priced at approximately RMB 1,099 (around US$160). Exciting AI tech isn’t reserved for the well-off anymore.

Rather than being tied to a single hardware manufacturer or AI model, the native-AI operating system can support multiple device makers, developers, and AI providers. With DroiClaw, users can:

Personalize AI avatars

Install skills

Create custom skills and scheduled tasks

Configure different models

As a result, the AI OS moves with you and your future AI needs.

Experience meets the next generation of AI

DroiClaw builds on more than a decade of operating-system development by Shanghai Droi Technology, founded in 2008. According to Shanghai Droi Technology, the company’s earlier FreemeOS platform has been deployed on more than 200 million devices worldwide and supported by a network of more than 1,000 hardware and software partners across Europe, India, Southeast Asia, Latin America, and other regions.

As the industry moves toward AI-native computing, operating systems are becoming central to how AI is delivered. The key is to combine:

System-level AI integration

An app-free Agentic OS vision

Hybrid on-device and cloud intelligence

An open ecosystem designed to broaden access to advanced AI

By doing this, DroiClaw offers a glimpse into how future smart terminals may evolve beyond today’s app-based experiences.