Your StackChan, your language model, your network — local-first firmware and server kit for the M5Stack StackChan. Voice, face, camera and MCP tools, with nothing sent to a cloud service.
Your StackChan, your language model, your network.
Firmware and a local server kit that turn an M5Stack StackChan into a voice assistant running entirely on your own machine: speech recognition, a local LLM, text-to-speech and the robot's face, head, lights and camera — with nothing sent to a cloud service.
By Drax and Claude. Built with Claude Code.
Working, on real hardware, daily. Voice, face, head, LED ring, camera, wake word, an on-screen settings menu and the privacy switches are all in use on the reference robot. Flash it from your browser — no toolchain, nothing to install — or build it yourself from this tree.
Verified on one StackChan. The model measurements cover two cards — a desktop RTX 4090 and a mobile RTX 5060 — and they rank models differently. If you have either a second robot or another card, a report from it is the most useful thing you could contribute — see CONTRIBUTING.md for what to send and what not to (no full-flash dumps: they carry your Wi-Fi password), and tools/model-bench if it is the GPU.
⚠️ Flashing writes to the device, and an interrupted write can brick it. Most failures are recoverable — the ESP32 has a bootloader that survives a bad app image, and M5Burner can restore a device that will not start. But a write interrupted at the wrong moment, by a cable nudged out, a power cut, a failing USB port or a hub that drops under load, can leave a robot that does not boot and does not come back easily. Those are circumstances outside anyone's control, including this project's.
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/DrakTheDude--stackchan-local-llm.html)
🦆🧠 Give your small robot a brain. Tell a Microduck, a Reachy Mini, a LeRobot arm or any ROS base what you want in plain language. An LLM (Claude, OpenAI, Gemini, Grok, or a local model via Ollama, vLL
MuhittinYilmazer/akanaSelf-hosted, local-first AI assistant server: swappable LLM providers, review-gated memory, an encrypted vault, and wake-word voice. No accounts, no telemetry.
Alex663028/pulse-agentPulse — Hermes-style self-improving AI agent. Reliability-first rebuild with evaluated skill self-evolution, multi-agent team orchestration, dialectic user modeling, and fully self-hosted default stac
Ambar-Gupta22/corvusNative C++17 AI agent runtime — local-first, embeddable, MCP-native. Tool-calling, memory, and an async agent loop that runs where Python can't: ROS2, game threads, drones, HFT.
cinderline/northcinderOpen-source MCP server for comparing products and asking the buyer before purchase.
vericle/heyclaw🦞 Headless real-time Python voice assistant powered by ElevenLabs, Gemini, Mem0, MCP tools, and local wake-word detection.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.