Chat to Build
Define device behavior through IM chat + dynamic Lua loading, with generated logic persisted as local deterministic rules
ESP-Claw is a Chat Coding AI agent framework for IoT devices. It defines device behavior through conversation and completes the full loop of sensing, reasoning, decision-making, and execution locally on Espressif chips.
Chat to Build
Define device behavior through IM chat + dynamic Lua loading, with generated logic persisted as local deterministic rules
Event-Driven
Devices report events proactively with millisecond-level response; if no rule matches, call LLM automatically with cloud-edge collaboration
Bi-Directional MCP
Devices act as both MCP Server and MCP Client, callable by external Agents while also calling external services proactively
Local Memory
Structured long-term memory with lightweight summary-tag retrieval, keeping privacy local
| Dimension | Traditional IoT (Cloud-Centric) | ESP-Claw (Edge AI) |
|---|---|---|
| Processing logic | Preset static rules (If-This-Then-That) | LLM dynamic decision-making + Lua deterministic rules |
| Execution engine | Rule engine | LLM + Lua + Router (three-tier event handling) |
| Control center | Cloud server | Edge node (ESP chip) |
| Device protocol | MQTT / Matter / proprietary SDK | MCP as a unified language + multi-protocol bridging |
| Memory management | Cloud data storage | Local structured memory (JSONL + tags) |
| Interaction model | App / control panel | IM chat (Telegram / WeChat / Feishu) |
| Intelligence | Preset automation | LLM + local rules (continuous evolution) |
This tutorial walks you from zero to a working ESP32-S3 breadboard build, then flashes the latest ESP-Claw firmware.

ESP-Claw is still under active development. Feel free to open an issue to report problems or request features. You can also share your ideas through our online survey (in Chinese).
Click here to view our TODO List (in Chinese), and vote for the features or issues you care about. That helps us prioritize them sooner.