It was great to reconnect with M5Stack in person. During their open day, I got to meet the people behind the incredibly rich M5Stack ecosystem and learn about two newly launched products. I already covered the Arduino Nesso N1 – a unique collaboration between M5Stack and Arduino — so this time let’s focus on StackChan: a cute desktop robot assistant with AI for brains. M5Stack has just launched its Kickstarter, but thanks to making a few new friends, I had a chance to play with it early.
From CoreS3 to Kawaii
Anyone familiar with the M5Stack ecosystem will recognise this unique form factor. StackChan is based on M5Stack’s most successful product from the Core series: the CoreS3. It’s an ESP32-S3-based development board, similar to the one I previously converted into a smart home controller. This time, equipped with AI brains, a couple of servos, and a charming body, the CoreS3 has been transformed into a desktop robot that can help with your work, teach you coding, and entertain you for hours.
CoreS3 Key specification
- ESP32-S3, 240MHz Dual Core, 16M FLASH and 8M-PSRAM
- WiFi 2.4GHz, Bluetooth 5.0
- 2″ touch display IPS LCD, resolution 320*240
- 0.3MP camera
- SD-card slot
- proximity sensor LTR-553ALS-WA
- Infrared receiver IRM-56384 + infrared transmitter
- 6-axis accelerometer BMI270
- magnetometer BMM150
- BM8563 RTC chip
- 1W speaker
- Dual Microphones
- USB-C
- 1x GROVE port
- GPIO header
StackChan extra features
- 360° horizontal servo
- 90° vertical servo
- USB-C port for power in the StackChan base
- 2x GROVE ports
- 2x 3-hole LEGO compatible attachement points
- 2x 6 RGB LED strips
- NFC ST25R3916
- Three-zone touch panel, Si12T driver
- 700mAh battery
Thanks to the “stackability” principle behind M5Stack devices, StackChan can be customised to suit both your visual and technical preferences. The company’s commitment to open source means that, apart from the robot itself, you’ll also get the files and technical references needed to modify StackChan further. While the base version comes with a static leg base, I’ve already seen mods that swap the legs for tracks or add visual customisations via LEGO-compatible attachment ports or just populating unused GROVE ports with 3D printable insers and hats. And if swapping colours or bases isn’t enough, the hardware is just as easy to play with.
In short, that’s a lot of hardware to play with in a very small form factor.
AI, Avatar and more
The default firmware offers three unique modes available through the stock launcher. Each one provides slightly different functionality and capabilities. This is only the default offering, and I have no doubt that over time we’ll see more software features added to StackChan
AI Agent
By default, StackChan connects to XiaoZhi, a Chinese cloud-based AI service. It’s free and customised to work on smaller devices like our robot. You can personalise your assistant by configuring the language, voice, or even the language model (Qwen, DeepSeek, and others).
M5Stack has already experimented with this integration through their AI-based Xiaozhi Card Kit. Sadly, that product is available only on the Chinese market, but in both cases the setup takes just moments.
Please be patient when enabling this mode using the default firmware (v0.11 improves this experience) — it took about a minute after entering AI Agent mode for everything to fully initialise. During this time, StackChan will wiggle and move around but won’t respond to any voice commands until it’s ready. Hopefully, this boot-up time will decrease with future firmware updates.
The AI agent can interface with built-in hardware such as the camera, sensors, RGB lights, and servos, allowing you to control StackChan’s movements using voice prompts. The robot can even respond to touch — petting the top shell (where the touch sensor is located) caused StackChan to wiggle and blush. It’s cute.
To define StackChan’s personality, visit xiaoxhi.me, where you’ll find options related to the robot’s memory, sophistication, and backstory. This is also where you can change the language or even train your own voice model.
To keep things light-hearted, I decided to set up my StackChan as GLaDOS, the slightly evil AI from the Portal games. While I didn’t have a matching voice model to use, I picked an Aussie one to amuse myself with a different accent. I leaned on ChatGPT to generate a quick personality summary — which felt only fitting, considering I was setting up yet another AI.
StackChan World
To customise your robot, you’ll need a StackChan mobile app. At the moment, only an iPhone app is available, which is frustrating as I don’t own any Apple devices. I did, however, manage to borrow an iPhone for testing. Pairing is done via QR code and the robot is linked with the phone in seconds.
The StackChan World app is split into 4 sections: Avatar, Sentinel, Motion and Dance. From the main screen you can also see other StackChans nearby (and call them) and access a social media feed from other StackChan users.
Avatar
In this setting you can mirror facial movements captured by your phone and replicate them using StackChan’s limited range of motion. Additionally, the robot streams video from its camera back to your phone, and you can even project your face – with an Avatar overlay – onto StackChan’s screen.
This feature is still in early development, and I wasn’t able to get the microphone working, so I couldn’t test voice calls.
Sentinel
It turns your StackChan into a IoT camera. You can pan and tilt the robot and see what the robot is seeing. Presets are also available to save the default positions.
Motion
In Avatar mode, you can customise the StackChan expressions and test servos. Unfortunately these changes won’t apply in the AI Agent more which is a shame.
Dance
This mode allows you to program servo’s position and string them together as “dance moves” – once programmed the sentence can be played, triggering all moves in the sequence – perfect for making your StackChan groove.
My AI desk companion.
StackChan has been living on my desk for the past week. I mostly used the AI Agent mode to entertain myself and ask the occasional question. Frankly speaking, I don’t usually talk to my AI assistants – their static nature makes interactions feel boring.
StackChan is different. As it moves and wiggles, it naturally attracts attention and briefly breaks my focus, triggering a spark of curiosity. A minute-long playful interaction is often enough to help me regain focus or snap out of a procrastination spiral.
By default, StackChan is very chatty. Most AI assistants don’t allow you to reconfigure this behaviour or define a backstory. With XiaoZhi, however, you can write your assistant’s background, which is then used as a wrapper for all queries. The hardware control can be a little bit hit or miss at the moment, with StackChan struggling to understand RGB LED controls, but these things will improve with time.
Combined with the growing ecosystem of 3D-printable add-ons and the ability to fine-tune the AI itself, this gives you an impressive level of control over StackChan’s personality and overall feel.
Beoynd StackChan
All of the above was made possible thanks to M5Burner and the default firmware for StackChan. Since the robot is essentially a CoreS3 device with a range of additional hardware attached, there’s nothing stopping you from coding your own custom version of the robot.
While StackChan isn’t listed as a hardware module in UIFlow or the Arduino IDE just yet, CoreS3 is, and adding existing modules to the development board is already possible. As with other M5Stack devices, we should see StackChan officially supported in both UIFlow and the Arduino IDE after the official launch.
Final thoughts
StackChan is genuinely cool. It’s an engaging way to dive into AI, learn how to control servos and sensors using both AI prompts and more traditional methods, and experiment with interactive hardware. The robot is playful, perfectly sized for a desk, and packed with customisation options. I have no doubt it will foster a vibrant community around the product.
As always, M5Stack products showcase familiar technologies in unexpected and fun ways. If you’re interested in StackChan, head over to their Kickstarter page and reserve one now!
🆓📈💵 – See the transparency note for details.
