01
Social-App Control
Issue commands to the gateway via WeChat, QQ, WeCom, DingTalk, Feishu and more — configure parameters, query status, control devices, assign tasks. No dedicated app or on-site operation needed.
Edge Computing Gateway
M150UB
Edge Computing Intelligent Gateway
Let AI Perceive the Real World · Social-App Control · Self-Growing Skills · Deploy Anywhere
2.0GHz
Processor Frequency
1TOPS
AI Computing Power
4W
Total Power
350g
Net Weight
-20~60°C
Operating Temp
Gallery
Overview
PCs can't enter the workshop. M150UB can.
Computers excel at office work but can't enter workshops, stand beside base stations, or be embedded in switchgear cabinets. The M150UB was born for exactly this — an edge computing intelligent gateway deployable at any field site, enabling AI to truly perceive and control the real world. With a built-in AI Agent engine, users can issue commands to the gateway via everyday social apps like WeChat, QQ, or DingTalk. More importantly, its software capabilities are alive — agents can autonomously create skills that grow continuously with your needs.
Social Control
Control Everything with Natural Language
No dedicated app, no on-site visit. Chat with the gateway's AI Agent via WeChat, QQ, WhatsApp, DingTalk and other everyday social apps — query devices, adjust parameters, and assign tasks using plain language.
Advantages
Core Advantages
02
AI Perceives the Real World
Built-in nanobot AI Agent engine with 1TOPS NPU, combined with industrial sensors and tri-network connectivity, takes AI beyond the screen — into workshops, base stations, and switchgear cabinets.
03
Self-Growing Skills
Unlike fixed-function legacy gateways, M150UB's agents can autonomously create and extend skills. Software capabilities are alive, continuously evolving with business needs.
04
Bidirectional Data Flow
Not just traditional collect-and-upload. The gateway connects via Ethernet, Wi-Fi, and 4G, proactively fetching internet data and calling cloud services for true edge-cloud collaboration.
05
Rich Industrial Interfaces
3× RS485 + 1× CAN FD + 2× DI/DO + Gigabit Ethernet + USB 3.0. One gateway covers all plant communication needs — from sensors to PLCs to actuators.
06
Industrial-Grade, Deploy Anywhere
IP40 protection, wide-temp operation, only 4W / 350g, DIN-rail mount. Deployable at any site that needs intelligence — the fundamental advantage of edge computing over traditional PCs.
Specifications
Specifications
General
ModelM150UB
ProcessorRK3568 Quad-Core Cortex-A55 @ 2.0GHz, 1TOPS NPU (INT8)
Memory / Storage2GB / 8GB or 4GB / 32GB (optional)
Operating SystemUbuntu
Software PlatformWinIF Designer / nanobot AI Agent / nanobot-web
SD CardBuilt-in, up to 64GB
Real-Time ClockBuilt-in RTC
Power
Rated Voltage12 ~ 24V DC
Voltage Range8 ~ 28V DC
Power Consumption4W
Inrush Current30A
Communication
Ethernet2 × 10/100/1000M Auto-Sensing (Shielded RJ45)
RS4853 Ports (COM1 / COM2 / COM3), Transient Withstand ±24V
CAN Bus1 × CAN FD (COM10), Transient Withstand ±24V
USB1 × USB 3.0 + 1 × USB Type-C
Wi-FiIEEE 802.11 b/g/n
4GChina Mobile / Unicom / Telecom (All Carriers)
DI2 Channels, Input Range 3.3 ~ 36V
DO2 Channels, Relay Output, Overload 5A 250VAC / 5A 30VDC
Environmental & Certifications
Ingress ProtectionIP40
Operating Temp-20°C ~ 60°C
Storage Temp-25°C ~ 70°C
Relative Humidity10% ~ 90% (Non-condensing)
Altitude≤ 3000 m
Vibration10 ~ 25Hz, 2G / 30min (X / Y / Z)
EMCSurge ±1KV, Burst ±2KV, ESD 4 / 8KV
CEEN61000-6-2:2005, EN61000-6-4:2007
RoHSCompliant
Mechanical
Dimensions133 × 120.5 × 51.5 mm
Net Weight350g
CoolingPassive (Fanless)
MountingDIN Rail
Software
Intelligent Software Platform
nanobot · Ultra-Light AI Agent Engine
nanobot is an ultra-lightweight AI Agent engine designed for edge devices. It's not just passively running preset programs — it's an autonomous agent that understands natural language, creates skills on its own, and extends capabilities on demand.
- Supports 15+ LLM providers: OpenAI, Claude, Gemini, DeepSeek, Ollama and more
- Native integration with WeChat, QQ, WeCom, DingTalk, Feishu, Telegram, Slack, Email
- Agents autonomously create Skills — capabilities grow dynamically with needs
- MCP tool invocation — connect external tool servers to extend physical-world control
- Layered memory system + Dream memory consolidation for continuous learning
- Bubblewrap security sandbox ensures isolated, controlled command execution
nanobot-web · Web Management Console
Zero code intrusion architecture — interacts with the core engine entirely through shell commands and config file parsing. All management features accessible via browser, no client software needed.
Network ManagementUnified management of ETH0 / WLAN0 / WWAN0, real-time connection monitoring
Agent TerminalVisual Agent interaction with real-time Markdown rendering, direct AI conversation
Service ManagementOne-click start/stop/restart, real-time telemetry (CPU / Memory / Network)
Config CenterCentralized API Key, model params, channel management with secure credential storage
Hardware MonitorReal-time TX/RX throughput, driver health status and hardware-level metrics
Architecture
Four-Layer Decoupled Architecture
Separated from hardware to application layer, balancing performance and flexibility. Unlike traditional gateways, every layer is designed for autonomous AI agent operation.
ApplicationAI Multi-Channel Push | OPCUA Server | Remote DB | WebVNC | API
Platformnanobot AI Engine | nanobot-web Console | WinIF Designer | MCP Toolchain
SystemUbuntu | Device Drivers | Network Stack | Security Sandbox
HardwareRK3568 + NPU | RS485 / CAN FD | Ethernet / WiFi / 4G | DI / DO
Bidirectional Data Flow
Data flow is no longer the traditional collect-and-upload pipeline. The AI Agent collects field sensor data for edge analysis while proactively fetching cloud information, calling external services, and responding to remote commands from social channels. Agents can autonomously create new skills based on operational experience, enabling continuous system evolution.
Scenarios
Application Scenarios
Production Line Intelligence
Collect PLC and sensor data in real-time via RS485 / CAN FD. AI Agent automatically detects equipment anomalies and pushes alerts — engineers reply on WeChat to remotely adjust parameters. Agents learn production patterns and create predictive maintenance skills.
Edge-Side Energy Analytics
Connect electricity, water, and gas meters for edge-side data aggregation and real-time analysis. Managers query energy reports via DingTalk or WeCom anytime. The gateway proactively fetches peak/valley pricing from grid platforms for AI-optimized scheduling.
HVAC & Lighting Coordination
Interface with BACnet / Modbus HVAC, lighting, and access control subsystems. Property managers send commands via QQ to remotely adjust temperature, switch lights, and check status. Agents learn usage patterns and create scene automation skills.
Remote Field Data Collection
Ideal for water quality monitoring, weather stations, and pollution source tracking in remote locations. Tri-network redundancy ensures uninterrupted field data transmission. AI Agent pushes anomaly alerts via Feishu, combining forecast data with field measurements for intelligent prediction.
Models
Models & Service
M150UB
2GB / 8GB
Built-in AI Agent for data collection, intelligent analysis, social control, and protocol forwarding
M150UB-Pro
4GB / 32GB
Extended memory and storage for multi-model inference, large-scale data caching, and complex skill expansion
Standard Accessories
M150UB unit × 1, 4G / Wi-Fi antenna × 2, DIN rail clip × 2, Quick Start Guide × 1
Warranty & Service
12-month quality warranty from date of shipment. Free repair or replacement for quality-related defects during warranty period. Includes remote technical support, firmware update delivery, and continuous nanobot AI engine version updates.