Making AI Assistants
Cheap Again
Deploy your own AI assistant on hardware that costs just one dollar. Seamlessly bridging the digital and physical worlds.
Deploy your own AI assistant on hardware that costs just one dollar. Seamlessly bridging the digital and physical worlds.
No cloud dependency. No Linux. Just pure C on bare-metal RTOS, with LLM superpowers.
No Linux. No Node.js. C99 on FreeRTOS and RT-Thread. Runs on MCUs with 400 KB RAM. Hardware costs one dollar.
Networked nodes discover each other, share tasks, and collaborate in real time. Each Claw is a neuron in a distributed brain.
The AI dynamically orchestrates GPIO, LCD, sensors, and more via function calling. No re-flashing needed.
Everything you need to build intelligent embedded agents.
Interactive conversation with Claude API over HTTP. Chat-first shell where direct input goes to AI.
LLM-driven hardware control via function calling. GPIO, system info, LCD drawing dynamically orchestrated.
320x240 RGB565 framebuffer with text, shapes, drawing primitives. AI draws on screen via tool calls.
Write once, run on FreeRTOS and RT-Thread. Clean claw_os.h hides all RTOS specifics.
Node discovery, heartbeat, and distributed task scheduling. Collaborative real-time execution.
Thread-safe message routing between services. Decoupled, modular architecture.
Short-term RAM ring buffer + long-term NVS Flash. Nodes learn and recall operation patterns.
Timer-driven execution and periodic automation. Recurring operations run autonomously.
Connect Feishu, DingTalk, QQ, Telegram as message channels. Control hardware from any chat.
Clean layered design with OSAL bridging multiple RTOS platforms.
AI drawing on LCD via Tool Use — the LLM controls hardware in real time.
Run on real hardware or emulate with QEMU. Current QEMU support covers Espressif fork targets and vexpress-a9; upstream STM32 boards are excluded for now because QEMU still lacks working Ethernet there.
| Feature | Description | Status |
|---|---|---|
| LLM Chat Engine | Interactive conversation with Claude API over HTTP | Done |
| Tool Use | LLM-driven hardware control via function calling | Done |
| LCD Graphics | 320x240 RGB565 framebuffer with AI drawing | Done |
| Chat-first Shell | UART REPL with AI integration and UTF-8 support | Done |
| OSAL | Write once, run on FreeRTOS and RT-Thread | Done |
| Gateway | Thread-safe message routing between services | Done |
| Networking | Ethernet + HTTP on QEMU; WiFi on real hardware | Done |
| Conversation Memory | RAM ring buffer + NVS Flash persistent storage | Done |
| Scheduled Tasks | Timer-driven task execution and periodic automation | Done |
| Swarm Intelligence | Node discovery, heartbeat, distributed scheduling | In Progress |
| Skill Memory | Learn and recall frequently used operation patterns | In Progress |
| IM Integrations | Feishu, DingTalk, QQ, Telegram as message channels | In Progress |
| Multi-Model API | Claude, GPT, Gemini, DeepSeek, and more | Planned |
| Web Config Portal | Built-in web page for runtime configuration | Planned |
| Claw Skill Provider | Serve as a skill for other Claws | Planned |
Flash firmware and open a serial terminal — directly from your browser. Requires Chrome or Edge.