PPIO Launches China’s First Agentic AI Infrastructure Service Platform, Equipping Agents with “Brains” and “Hands”

PPIO, a leading Chinese independent distributed cloud computing service provider, showcased its latest innovations at the 2025 World Artificial Intelligence Conference (WAIC) held from July 26th to 29th. At the event, PPIO Co-founder and CEO Yao Xin unveiled the company’s Agentic AI infrastructure (Infra) service platform—the first of its kind in China—designed to accelerate the development and large-scale deployment of AI Agents.

Addressing the specific development needs of AI Agents, PPIO has introduced its AI Agent Platform, available in both a General and Enterprise edition. The General Edition leverages a cost-effective distributed GPU cloud infrastructure. It features China’s first Agent Sandbox compatible with the E2B interface, along with model services optimized for Agent construction. This provides individual developers and small to medium-sized businesses with a high-performance, efficient, and affordable AI Agent platform. The Enterprise Edition is tailored to the diverse Agent requirements of medium to large enterprises. It integrates PPIO’s GPU cloud services, industry applications, and open ecosystem. Key differentiating advantages include long-term memory, controllable processes, full-domain operation, business linkage, and multi-modal collaboration. These features enable the creation of intelligent, closed-loop decision-making and execution systems, pushing beyond the limits of traditional AI. PPIO aims to deliver “intelligent agents” capable of understanding, seeing, thinking, and acting, empowering various industries with next-generation AI.

PPIO Leads the Way with Agentic AI Infra: Introducing China’s First Agent Sandbox

Artificial intelligence is rapidly evolving towards Agentic AI. Agent applications, capable of autonomously perceiving, planning, and executing complex tasks, require powerful Large Language Models (LLMs) as a “brain” to enhance understanding and decision-making. They also need agile “hands and feet” to execute precise API calls and tool operations, increasing flexibility and versatility. This technological shift necessitates a comprehensive infrastructure service platform tailored for the Agentic AI era, providing systematic support for Agent application development. “As the first domestic infrastructure service provider focused on Agentic AI, PPIO is dedicated to a comprehensive strategy—from China’s first Agent Sandbox to model services to a full-fledged AI Agent platform—to provide developers and enterprises with a full-stack Agent infrastructure service,” says Yao Xin.

Unlike traditional programming approaches, the code generated by Agents during task execution carries inherent uncertainty regarding quality. Potentially harmful code could jeopardize system security and impact other users. Therefore, creating a secure, isolated environment for Agent execution is crucial. PPIO’s China-first Agent Sandbox, compatible with the E2B interface, addresses this challenge. This cloud-based execution environment is specifically designed for Agent tasks, providing Agents with secure, reliable, efficient, and agile “hands and feet.” The Sandbox supports dynamic calls to a variety of tools, including Browser use, Computer use, MCP, RAG, and Search. PPIO’s Agent Sandbox is already integrated with prominent open-source projects, like Camel AI, OpenManus, and Dify.

Built on Firecracker MicroVM, the PPIO Agent Sandbox boasts robust security isolation, millisecond-level startup speeds, and high concurrency creation capabilities. With no pre-deployment needed, it’s ready for immediate use, ensuring that all Agent operations remain “restricted and controlled.” Stringent security isolation enables complete isolation between different Agent Sandbox environments. When multiple tasks run concurrently, each can operate in an independent environment, preventing data leaks and resource contention at the root. Ultra-fast startup times of less than 200ms, significantly faster than traditional virtual machine sandboxes that can take minutes or longer, allow for instant creation of sandbox environments, code execution, and previewing of results, dramatically improving developer workflow. Furthermore, the PPIO Agent Sandbox supports rapid, concurrent creation of thousands of sandbox instances to meet the high concurrency requirements of various business scenarios, accelerating result delivery and ensuring a superior user experience.

Moreover, compared to traditional centralized cloud services, PPIO’s globally distributed computing network enables Agent Sandbox resource pricing that is up to 50% lower than E2B’s official rates, further optimizing Agent Sandbox costs for enterprises. Customers also gain access to advanced features such as 24-hour session persistence, high concurrency, and custom specifications without monthly subscription fees. PPIO allows businesses to reduce their overall operating costs by more than 50% while maintaining full E2B compatibility, providing a faster, safer, and more economical foundation for large-scale Agent application deployment.

Model Services: Custom-Built for Agent Development, Supporting Ultra-Long Context Output

Building on a secure and controllable execution environment, PPIO’s model services offer immediate access to and optimization of leading open-source Large Language Models, providing the “brains” needed for Agent decision-making. PPIO’s Model Services support quick access, flexible deployment, and efficient utilization of hundreds of mainstream open-source and customized AI models. This empowers developers and enterprises to rapidly build Agent applications, fulfilling the core requirements of Agent tasks like planning, memory, and execution.

PPIO offers domestic mainstream inference models, such as DeepSeek R1, Qwen3, and MiniMax M1, to empower the Agent’s planning ability by breaking down complex tasks into multiple steps. To meet Agent’s memory needs, PPIO spearheaded the industry by upgrading the DeepSeek context window to 160K, with maximum output extending to 160K. This supports long output applications for multi-turn ultra-long conversations and in-depth Agent analyses. PPIO also supports code models such as Kimi-K2 and Qwen3 Coder to provide robust autonomous programming, tool calling, and mathematical reasoning capabilities, catering to various Agent application requirements through its extensive multi-modal model support covering text, voice, video, and images.

Additionally, PPIO leverages collaborative innovation across algorithms, systems, and hardware in its large model inference acceleration technology. With its self-developed inference acceleration engine, PPIO has deeply optimized core technologies such as PD separation, FlashMLA, DeepEP, DeepGEMM, EPLB, and MicroBatch Overlap, completely removing the “hardware constraints” on Agent development. This results in a >10x reduction in model inference costs, achieving a dynamic balance between inference efficiency and resource utilization.

“Brain-Eye-Ear-Mouth-Hand” Collaboration: Creating “Digital Labor” for Enterprise Intelligence

The rapid emergence of Agents presents a new opportunity for digital transformation of large enterprises in various industries. However, due to the sensitivity of core business data, private deployment is becoming the common choice for Agent implementation among these enterprises. Building industry-specific Agents and achieving private deployment not only faces core challenges such as “difficult computing adaptation, difficult model integration, and difficult cost control”, but also requires deep integration of industry knowledge and data to achieve “industry-knowing and professional” precise empowerment.

To address the needs and pain points of medium-to-large enterprises looking for industry-specific Agents, PPIO has launched the Enterprise version of its AI Agent Platform. Utilizing “brain-eye-ear-mouth-hand” collaboration, it creates intelligent decision-making to execution closed loops, remodeling automation with “digital labor“. The platform covers multiple fields such as education, industry, and enterprise scenarios, integrating infrastructure, foundation models, a model factory, a knowledge base brain, and intelligent agent orchestration capabilities. With an intelligent foundation driving enterprise efficiency upgrades and injecting “full-chain intelligent dynamics” into digital transformations, it enables automation to upgrade from a tool to a business growth engine, empowering enterprises to efficiently implement intelligent applications.

Currently, the Enterprise version of PPIO’s AI Agent Platform has practical applications: Based on the AI Agent Platform, PPIO has created a full-scenario smart education solution for an international school covering management, teaching, learning, and research, throughout all stages from pre-class preparation, classroom assignments, learning analysis, psychological monitoring, and home-school interaction matching all educational stages inside and outside of campus. The teacher can utilize AI’s efficiency to consolidate bilingual resources and streamline preparation time significantly reducing it by over 50% daily. The platform links the teacher-student-parent triangle which compresses information feedback time from one day to just 1.5 hours. During the recent monthly exams, the average scores of three grades improved by approximately 10%, directly demonstrating the value of the platform.

“From AI Infra to Agentic AI Infra, the PPIO AI Agent Platform not only provides Agent developers and businesses with comprehensive support from underlying computing power to upper-layer applications which lowers the technology barrier and development costs, but also ensures a secure, efficient sandbox environment, rich and diverse model resources, and extremely cost-effective computing services. This establishes a solid foundation for the large-scale deployment of Agents, enabling them to move from technological exploration to a new phase of value creation,” Yao Xin said.

Looking ahead, by continuously improving its platform service capabilities, PPIO will help more companies and developers seize the first-mover advantage in the Agentic AI era, accelerate the birth of innovative applications, transform the working model and production processes of many industries, unleash new productivity, and lay a solid foundation in creating a more intelligent, efficient, and innovative future industrial ecosystem.