● Open source · AI + IoT

Agentic AI
on real devices
with TuyaOpen

Build once, deploy across chips.

TuyaOpen powers next-gen AI-agent hardware: it supports gear (Tuya T-Series Wi-Fi/BT MCUs, Pi, ESP32s) via a flexible cross-platform C/C++ SDK, pairs with Tuya Cloud multimodal AI, integrates top models (ChatGPT, Gemini, Qwen, Doubao, and more), and streamlines open AI-IoT ecosystem building.

Quick Start →About TuyaOpen Applications GitHub

terminal

git clone https://github.com/tuya/TuyaOpen.git
cd TuyaOpen
. ./export.sh

tos.py check
cd apps/tuya_cloud/switch_demo
tos.py config choice
tos.py build
tos.py flash

Apache 2.0

Open source license

C/C++ SDK

Cross-platform from MCU to SoC

Tuya Cloud

IoT + multimodal AI

Community

GitHub & Discord · 8000+ active developers

Why TuyaOpen

Make development straightforward

Layered SDK, multimodal AI, and cloud-ready building blocks.

Layered SDK you can grow on

TKL (hardware abstraction), TAL (OS/device abstraction), libraries, services, and applications—develop once, deploy everywhere with reusable building blocks.

Multimodal AI on the edge

Speech (ASR, KWS, TTS, STT), vision, and sensor-based features. Integrate leading LLMs and platforms (DeepSeek, ChatGPT, Claude, Gemini, and more).

Cloud-connected and secure

Connect to Tuya Cloud for remote control, monitoring, and OTA updates. Built-in security, device authentication, and data encryption.

From prototype to production

Production-ready architecture from day one: reusable layers, stable connectivity, security-by-design, and scalable cloud integration so teams can move from proof-of-concept to shipped products with less rework.

Who is it for?

Built for builders at every stage

From first prototype to scaled shipment, TuyaOpen helps you build bold IoT and agentic hardware faster.

Students and learners

Turn ideas into working devices with a practical stack and clear workflows. Learn by building real AI+IoT products, not toy examples.

Makers and IoT enthusiasts

Move from hackathon concepts to polished demos quickly. Reusable SDK layers and ready integrations help you ship cool hardware with less glue code.

AI agent hardware creators

Design voice-first, multimodal products with agent workflows, tools, and cloud orchestration that map cleanly to real devices.

Commercial product teams

Adopt a production-oriented architecture with security, OTA, and scalable cloud capabilities to reduce risk and speed time-to-market.

Developer workflow

From clone to flash

Write once, target every chip class—from embedded MCUs to application SoCs—with one straightforward workflow. Develop on Windows, macOS, or Linux: the same tos.py toolchain on the OS you already use. Clone the repo, run export at the root to bring the SDK online, then follow the commands below. Host prerequisites and Windows export scripts: Environment setup.

terminal

git clone https://github.com/tuya/TuyaOpen.git
cd TuyaOpen
. ./export.sh

tos.py check
cd apps/tuya_cloud/switch_demo

tos.py config choice
tos.py build
tos.py flash

Get started in three steps

1. TuyaOpen license

Get a TuyaOpen dedicated license (UUID + AuthKey) for Tuya Cloud access. See Get Started for acquisition and writing options.

Get TuyaOpen license →

2. Clone, init SDK, and pick a board

Clone TuyaOpen from GitHub or Gitee, cd into the repo, and run the export script (. ./export.sh on Linux/macOS, or export.ps1 / export.bat on Windows) to initialize the tos.py environment. Run tos.py check, then in your app folder use tos.py config choice for the board.

Step-by-Step Guide →tos.py guide

3. Build and flash

From the application directory, run tos.py build and tos.py flash. Use tos.py monitor when you need serial logs or authorization steps.

Device-Cloud One pager

On the edge you ship firmware with the TuyaOpen C/C++ SDK; in the cloud you use the Tuya Cloud developer platform for multimodal AI. Text, speech, vision, and sensor data move between device and cloud so you can build next-gen AI-agent hardware and open smart ecosystems.

Hardware & TuyaOpenSDK

Targets from Tuya T-Series MCUs to Raspberry Pi, Rockchip and Linux SoCs, and ESP32—one RTOS-like, modular stack.
Rich peripherals and drivers; build generic IoT and AI-agent applications with familiar Win / macOS / Linux toolchains.
Device-side building blocks pair with cloud AI APIs for Wi-Fi IoT and multimodal features.

Tuya Cloud platform

Cloud-based multimodal AI: low-latency speech and vision, zero/low-code integration, and drag-and-drop AI agent workflows.
Long- and short-term memory, custom MCP servers, emotion-aware TTS/STT, and MiniApp app panels.
Works with leading models (ChatGPT, Gemini, Claude, Qwen, DeepSeek, and more) and the Tuya Smart hardware ecosystem.

Overview: TuyaOpen hardware development (left) and Tuya Cloud (right). Click to enlarge

Supported chipset platforms

From coin-cell MCUs to Linux-class SoCs—pick the silicon that matches your power budget, latency goals, and AI topology. Host tooling on Windows, macOS, and Linux. See About TuyaOpen for the full matrix.

MCU

Ultra-lightweight, low-cost, and power-efficient—built for always-on IoT. Stream sensor and media data to Tuya Cloud; multimodal AI is processed in the cloud so your device stays lean.

Tuya T5MultimodalRecommended

Tuya T2

Tuya T3

ESP32-C3

ESP32-S3

ESP32-C6

SoC

More headroom for edge AI: run richer models locally, then fuse with Tuya Cloud AI for scale, orchestration, and the latest LLMs—one stack from device to cloud.

Raspberry Pi

Linux

Ubuntu

Rockchip

Demos to try

Start from production-oriented demos. Pick a target, open the source, and follow the guide to run it on your board in minutes.

switch_demo

Baseline IoT firmware for networking, pairing, cloud control, and OTA across supported targets. A simple and straightforward IoT on/off example with a controllable app panel.

Code Guide

your_chat_bot

Voice-first AI chat demo with agent interaction, tuned for T5AI, ESP32-S3, and Raspberry Pi 4/5 class devices. Supports multimodal input and skill extension, ideal for personal assistants and conversational experiences.

Code Guide

duo_eye_mood

Dual-eye expressive demo with eye and mood visuals, plus conversational interaction on T5AI.

Code Guide

Rich Peripheral Examples

From hardware protocols and OS-level programming to LVGL GUI and generic examples, this is a practical building-block set to get started fast.

Code

Tuya T5 : Cost-effective for Agentic AI

Tuya T5 chip/module is a high-performance embedded Wi-Fi 6 + Bluetooth 5.4 dual-mode communication module, embedded with ARMv8-M Star (M33F) processor and a main frequency up to 480MHz. The chip is purpose-built for multimodal AI interaction scenarios with audio, video, and display enablement. Rich GPIO resources accelerate integration, and built-in Wi-Fi 6 (2.4 GHz) plus BLE connectivity simplifies product bring-up.

Learn more →

Experimental feature

Now Develop agentic AI hardware with Arduino IDE.

Prototype on T5 with a familiar Arduino workflow—board support and libraries integrated with TuyaOpen so you can move fast on agentic AI hardware.

Learn more →

Multimodal AI

Tuya AI: Zero-Code Cloud Platform

Tuya AI: Multimodal edge-cloud intelligence with edge AI inference and a cloud agent hub, enabling access to leading AI models (DeepSeek, ChatGPT, Claude, Gemini) and cross-modal functions including voice/text interaction and image/video generation for edge devices. On the Tuya Cloud platform, you can build zero-code agents and customize IoT behaviors without rebuilding firmware.

Agentic Hardware Control
Memory RAG
Agent Workflow
Agent Prompting
ASR
Skills (Weather, IoT Control, and Third-party)
Model Context Protocols (MCPs)

Learn more →

Tuya AI platform screenshot: dashboard overview.

Applications & use cases

Applications and use cases

Reference designs, products, and industry patterns with TuyaOpen: devices above, plus smart buildings, industrial IoT, and retail—IoT connectivity and agentic AI on one stack.

AI Emoji Desktop Robot

A desktop companion that blends expressive on-screen faces with AI listening and understanding. Rich emoji-style reactions meet gesture and voice control: speak or wave, and the display tracks smoothly for a responsive, playful presence.

Emoji desktop robot with expressive display.

Developer Story

Cattle Tracking with AI Agents

HTX Studio: Bringing Smart Farming to Remote Villages

When HTX Studio stepped into remote mountain villages, they saw farmers trekking rough slopes daily to find cattle, getting hurt again and again, while regular (4G Cellular) trackers simply lost signal in blind mountain terrain. Moved to help, the team leveraged TuyaOpen open-source capabilities to integrate long-range LoRa connectivity with local-dialect AI voice interaction. They quickly built a complete smart grazing system, skipping heavy low-level bottlenecks. With ready device-cloud and AI toolchains, they turned empathy into simple, reliable smart technology for farmers deep in the mountains. Conversational AI cattle tracking makes daily interaction easier, with almost no learning curve for elderly users.