DuckyClaw: A hardware oriented OpenClaw

Hey makers!

If you’ve played with OpenClaw, you’ve definitely asked: Can this AI agent run on my real embedded hardware, not just a desktop? We built DuckyClaw to fix that—no hacks, no workarounds, just a native hardware AI agent stack for the Claw experience on Edge device.

Core Features

I built this to turn AI agents from desktop chatbots into physical automation workhorses—here’s what it actually does for you:

✅ Cross-hardware native: T5-AI Board, ESP32-S3, Raspberry Pi 4/5, ARM Linux, x64 PCs. One C codebase for all, no re-coding per board.

✅ Edge + Cloud hybrid: Local agent loop for speed/privacy; Tuya Cloud unlocks multi-agent, RAG, ASR/TTS, web search—all with one TuyaOpen key.

✅ Native IoT control: Directly control real Tuya smart devices (lights, plugs, sensors, cameras) — no virtual sims.

✅ Custom hardware skills: Plug in cameras, sensors, displays, motors—fully integrated into the agent loop.

✅ Unified messaging: Telegram, Discord, Feishu in one gateway; on-board ASR voice input (no extra apps).

✅ Proactive agent loop: Background running, health checks, autonomous triggers—no constant manual wake-up.

✅ Persistent memory: IoT memory learns your preferences, keeps context across reboots.

Why It Crushes Other Claw Projects

Other forks lock you to one board or bloat MCUs with Node.js. DuckyClaw uses TuyaOpen C SDK for bare-metal performance, and taps into a real global IoT ecosystem—so you control actual devices

Architecture

(Left)Input: Messaging Apps, Voice Input

(Mid)Main Hardware Agent Loop: Gateway, Memory, Hardware Skills and Tools, Agent Loop and Heartbeat mechanism

(Right)Cloud: LLMs and zero-coding expansion features like MCPs and more Skills available.

Quick Start: 3 Steps to Get DuckyClaw Running

No complex setup—get up and running in 15 minutes, even if you’re new to embedded AI:

Step1:Grab Your Hardware: Start with a compatible board (we recommend Tuya T5-AI Board for out-of-the-box IoT support, or ESP32-S3/Raspberry Pi 4/5).

Step2:Get Developer License: One License Key to access all LLM/TTS/MCP/Skill Services

Doc: https://tuyaopen.ai/docs/faqs/get-developer-license

Step3:Follow The Quick Start Guide for Each Platform:

• Tuya T5 MCU Guide: DuckyClaw Quick Start (T5-AI)
• Raspberry Pi Guide: DuckyClaw Quick Start (ESP32-S3)
• ESP32 MCU Guide: DuckyClaw Quick Start (raspberry pi 5)

Advanced Topics

Create your own AI Agent callable Function Skills: https://tuyaopen.ai/docs/duckyclaw/custom-device-mcp

Use the `AI_MCP_TOOL_ADD` macro to create and register the tool. Define input properties with the MCP_PROP_* macros so the agent knows what parameters to pass.

TUYA_CALL_ERR_RETURN(AI_MCP_TOOL_ADD(
    "sensor_temperature_read",
    "Read the current temperature from the onboard sensor in degrees Celsius.",
    my_sensor_read_cb,
    NULL
));

Overview and Demo Video:

We’ve got your back with resources to avoid roadblocks:

Official Docs: Step-by-step hardware setup, SDK guides, and skill development tutorials — tuyaopen.ai/duckyclaw

GitHub Repo: Pre-configured project files, bug fixes, and community contributions — https://github.com/tuya/DuckyClaw

Maker Community: Join our Discord (link in GitHub) to ask questions, share your projects, and get help from the DuckyClaw team.