Engineered specifically for continuous 24/7 homelab operation, containerized AI agent workflows, and compact edge virtualization clusters.
Ultra-low TDP edge node optimized for quantized SLMs, Home Assistant automations & lightweight agents.
High-throughput multi-core NPU workstation built for local RAG pipelines and Proxmox virtualization.
Dual 2.5GbE LAN, expandable NVMe RAID storage, and continuous agent swarm execution.
Enterprise-grade thermal stability designed for autonomous micro-services and private local LLM servers.
Home lab enthusiasts, DevOps engineers, and self-hosting architects are transitioning from idle x86 mini servers to autonomous, AI-accelerated Agent PCs.
Unlike conventional homelab CPUs that throttle under matrix operations, modern Agent PCs feature integrated Neural Processing Units (NPUs) offering 45+ TOPS. This enables sustained 24/7 background LLM inference and vision embeddings at a fraction of the thermal and electrical footprint.
Running autonomous agentic loops (e.g., CrewAI, AutoGen, LangGraph) locally eliminates recurring API subscription costs and prevents sensitive personal or corporate data from ever traversing third-party public cloud endpoints.
With native dual 2.5G/10G Ethernet controllers and PCIe 4.0/5.0 interconnects, Agent PCs seamlessly integrate into K3s, Docker Swarm, and Proxmox VE environments to deliver sub-millisecond inter-agent communication across local distributed networks.
Explore how engineers and tech-forward enterprises leverage Agent PCs across complex self-hosted environments.
By clustering high-speed NVMe storage and high-bandwidth DDR5 memory, an Agent PC serves as an on-premise Retrieval-Augmented Generation (RAG) powerhouse. It continuously monitors self-hosted Nextcloud or Paperless-ngx instances, generates vector embeddings using lightweight embedding models, stores them in Milvus/Qdrant databases, and empowers local agents to answer complex queries with zero external data leaks.
Integrate an Agent PC into your local software development pipeline. Running localized coding assistants (such as DeepSeek-Coder, CodeLlama, or StarCoder via Ollama / vLLM), developers receive autocomplete, unit test generation, and automated code reviews within their private Gitea/GitLab repositories without latency or cloud token metering.
Combined with Frigate NVR and Home Assistant, the Agent PC's integrated NPU and GPU accelerate computer vision tasks. Instead of simple motion triggers, multi-modal Vision-Language Models (VLMs) analyze RTSP camera streams locally, describing complex occurrences (e.g., "delivery driver placed a package behind the pillar") and triggering intelligent automation routines.
Deploy independent agent swarms dedicated to market monitoring, automated RSS digest curation, vulnerability scanning, and homelab health diagnostics. Using frameworks like LangGraph and CrewAI running in containerized microservices, these agents collaborate asynchronously around the clock with near-silent acoustic output.
Deploying multiple containerized agents alongside virtualization hypervisors requires strict hardware reliability standards. Agent PCs engineered by industry leaders incorporate industrial-grade copper heat pipes, dual ball-bearing cooling fans, multi-stage power delivery, and extensive BIOS support for auto-power-on (AC Power Loss Recovery), PXE boot, and PCIe bifurcation.
How decentralized inference, model quantization, and hybrid architectures are shaping the next generation of intelligent edge deployments.
Founded in 2013, Shenzhen CYX Industrial Co., Ltd. is a wholly-owned subsidiary of CTONE Technology Group, specializing in AI PCs, Mini PCs, laptops, AI NAS, and intelligent computing products. With over 1,500 employees and manufacturing facilities in Shenzhen and Mexico, the company serves customers in more than 180 countries and regions worldwide. In 2025, the Group achieved annual revenue of RMB 5 billion.
As part of the Group’s AI ecosystem, KZHY focuses on enterprise AI applications and ecosystem integration, providing end-to-end solutions spanning AI terminals, intelligent computing servers, and data centers across cloud, edge, device, and computing scenarios.
In 2026, the Group launched its "All in AI" Strategy and established strategic partnerships with industry leaders including Intel, AMD, Alibaba Cloud, SenseTime, mThreads, and hello-tech. Looking ahead, the Group will continue to embrace an open ecosystem philosophy and welcomes more partners to join us in shaping the future of AI together.


Select from our comprehensive range of specialized AI hardware platforms tailored for homelab hypervisors, edge clusters, and enterprise edge computing.
Energy-efficient platform designed for compact micro servers and silent edge inferencing.
High-performance AI workstation for multi-agent swarm control and heavy containerized loads.
Optimized for Proxmox VE clustering, containerized microservices, and private RAG vector DBs.
Dual multi-gigabit Ethernet interfaces with accelerated NPU pipeline for sovereign local LLMs.
Combines high-density storage management with intelligent agent file sorting and semantic indexing.
Ruggedized thermal enclosure engineered for non-stop industrial automation and vision monitoring.
Centralized brain for complex Home Assistant logic, local voice processing, and camera VLM feeds.
Automated local code review, CI/CD pipeline triggers, and continuous security compliance scans.