Mr.Z AI chatbot
A voice-first AI assistant that runs on your device — talk to it out loud, get spoken answers.
Mr.Z AI chatbot turns your Attaky into a voice assistant you talk to out loud. Ask a question, hold a back-and-forth conversation, and hear spoken replies — using the device's own microphone and speaker.
This firmware supports all Agent Series Complete Builds.
What you can do with it
- Talk to it and hear spoken answers
- Hold a natural back-and-forth conversation
- Use it hands-free with the built-in mic and speaker
What you need
| Module | |
|---|---|
| Required | Core: ESP32 1.0 + Expand: Media I/O 1.0 (mic + speaker + microSD) |
| Optional | Power: Standard Cell 1.0 (battery) |
Already have an Agent Series build? It's ready to flash — Mr.Z runs on all of them.
Install it
Firmware is installed from the Attaky Firmware Store with a one-click web flasher — no toolchain to set up.
Open the Attaky Firmware Store in a supported browser (Chrome or Edge) and connect your Attaky over USB-C.
Pick Mr.Z AI chatbot and click flash. Keep the device connected until it finishes.
The device reboots into Mr.Z. Power on and start talking.
Don't unplug while flashing. Make sure you pick the firmware that matches your hardware — Mr.Z needs the Media I/O module for its mic and speaker.
Flash manually (advanced)
Mr.Z is built with ESP-IDF and ships as several images. The OTA and speech-model slots sit at addresses defined by the partition table, so read that first:
| File | Role | Flash address |
|---|---|---|
bootloader.bin | Bootloader | 0x0 |
partitions.bin | Partition table | 0x8000 |
ota-data.bin | OTA data | per partition table |
firmware.bin | Application | per partition table (often 0x10000) |
speech-models.bin | Offline speech models | per partition table |
# Inspect the partition layout first
esptool.py --chip esp32s3 image_info partitions.bin
# Then flash with the addresses it reports
esptool.py --chip esp32s3 write_flash \
0x0 bootloader.bin \
0x8000 partitions.bin \
<addr> ota-data.bin \
<addr> firmware.bin \
<addr> speech-models.binBuilt with ESP-IDF — the OTA and speech-model addresses come from the partition table, not the standard Arduino defaults.
First steps
Power on and start talking — Mr.Z listens on the built-in microphone and answers through the speaker. Add the Standard Cell battery module to use it untethered.
Versions
The Attaky build tracks the upstream release directly.
| Version | Status |
|---|---|
| v1.6.0 | Current |
Upstream & license
Mr.Z is Attaky's port of an open-source assistant project, adapted to Attaky hardware and made available to flash.
- Upstream: 78/xiaozhi-esp32 v1.6.0
- License: MIT — © 2025 Shenzhen Xinzhi Future Technology Co., Ltd. and project contributors
- Attaky build: a port to Attaky hardware, with the upstream copyright and license preserved
Known limitations
Mr.Z is evolving quickly, and Attaky hardware support keeps improving with each release. Run into something that doesn't work? See Troubleshooting & FAQ.