I built an AI Tamagotchi before Meta could

Turning a Tiny ESP32 Pendant into an AI Tamagotchi

Meta’s upcoming Muse Charm promises a cute, pocket‑sized AI companion that looks like a modern Tamagotchi. While the official device is still shipping, I bought a $50 Waveshare ESP32‑S3 1.75‑inch touchscreen pendant and used Claude’s code‑generation abilities to give it the same personality – and far more functionality – without any prior programming experience.

Getting Started: From Box to Connected Device

The pendant arrived as a simple development board with a tiny touchscreen, a push‑to‑talk button, and a USB‑C port. Setting it up was straightforward:

1. Plug the board into a PC and install the provided driver.
2. Clone Meta’s Muse Gadget repository from GitHub.
3. Pair the device with the Muse phone app and upload Meta’s API keys.

At first, Muse interpreted my voice as audio‑to‑text requests, responded slowly, and displayed only a few words on the screen. The default avatar was a single static “Jolly” face, and there was no integration with smart‑home devices.

Claude Steps In: Filling the Gaps

A few prompts to Claude (the Anthropic model) transformed the pendant into a usable companion:

• Conversation History: Claude generated a simple UI that stores each exchange, letting me swipe right to read full responses.
• Voice Output: By integrating a free British voice from ElevenLabs, the pendant now speaks answers aloud, making interactions hands‑free.
• Remote Camera Viewfinder: Claude added a Bluetooth‑based preview from my phone’s camera and a one‑tap shutter button. The phone now sends a high‑quality photo back to the pendant, even when I’m away from Wi‑Fi.
• Pocket‑Watch Face: A tiny analog clock with three hands was coded in under 15 minutes, turning the device into a functional watch.

Adding Personality: Avatars and a Tamagotchi Minigame

Meta’s demo only offered a single avatar. I asked Claude to create a set of four distinct avatars, each with its own idle animation. Then we built a Tamagochi‑style minigame:

• The pet has hunger, boredom, and cleanliness meters.
• Feeding it produces a crunchy sound and a mess that dirties the avatar.
• Over‑playing tires the pet, requiring rest.
• Levels increase the complexity of each meter, keeping the game engaging.

All of this was achieved with a handful of Claude‑generated scripts that run on the ESP32’s MicroPython environment.

Practical Benefits and Limitations

What works well:

• Voice interaction is now audible and responsive (limited only by Meta’s server latency).
• The remote camera feature works over Bluetooth, so I can capture photos from across the room without a Wi‑Fi connection.
• The pocket‑watch UI provides a genuine secondary function beyond the AI companion.

What still feels rough:

• Meta’s backend can be inconsistent; occasional lag is noticeable.
• Security concerns remain – I keep the pendant on a guest Wi‑Fi network and avoid storing personal data on the device.
• The code is experimental; without a developer background, troubleshooting can be intimidating.

Should You Build Your Own or Wait for the Muse Charm?

If you enjoy tinkering and want a device that does more than the commercial Muse Charm, a cheap ESP32‑S3 board plus an LLM like Claude offers a powerful playground. You gain full control over avatars, voice, smart‑home integration, and even custom games. However, the process still requires patience, a willingness to iterate on generated code, and a cautious approach to privacy.

For most users, the official Muse Charm will likely provide a polished, out‑of‑the‑box experience with tighter security. But for hobbyists who love to hack, the development kit demonstrates how quickly a modest hardware platform can become a personalized AI companion.

---

This article is a practical recount of building an AI Tamagotchi using publicly available tools. No proprietary specifications or unreleased product details are disclosed.