Meta has made the code public that lets you build your own Muse devices. The company is sharing the code behind its AI agent, and it wants you to attach it to almost anything you own.
The announcement comes with a clear promise: you can build these devices yourself. Meta describes them as “open source devices you build yourself,” and the company points to several projects that show what’s possible. One example involves attaching Muse to a color E Ink display so it shows reminders. Another idea puts Muse on a small touchscreen device that looks like a DIY version of the Muse Charm. A third project adds Muse to an HDMI stick, letting you display the agent on a bigger screen.
What the Code Does
The code is designed to work with common hardware. Meta suggests programming an off-the-shelf ESP32 board or setting up a Raspberry Pi with its SDKs. From there, you connect Muse to displays, buttons, sensors, actuators, and anything else you have lying around on your workbench.
The company warns that this is early-stage work. It cautions builders to “Proceed at your own risk!” That warning sits alongside the encouragement to experiment, and it reflects the nature of open-source development: the code is shared, but the company is not guaranteeing that everything works perfectly out of the box.
The Home Link Gadget
Meta is not just sharing code. It is also giving away a physical device it built itself. The Muse Home Link connects to community-built skills, letting the AI agent turn on a light, control a TV, send a document to a printer, and more, depending on your setup.
“Muse gadgets are open source devices you build yourself.”
The company made 5,000 of these gadgets, according to Nat Friedman of Meta Superintelligence Labs. You can join a waitlist for the Muse Home Link now, with shipping expected sometime this month.
Why This Matters
Open-sourcing the code changes the relationship between a company and its users. Instead of buying a finished product, you can now build the thing yourself. That shift opens up a lot of possibilities.
The ESP32 board and the Raspberry Pi are both off-the-shelf parts, meaning you do not need to order custom components to get started. The SDKs provide the bridge between those hardware platforms and the Muse software. By making those tools available, Meta is lowering the barrier to entry for anyone who wants to experiment with AI agents in their home.
What You Can Build
The examples Meta gives are practical and varied. A reminder display could sit on your desk and show you messages without needing a full computer. A small touchscreen device could serve as a personal assistant for quick tasks. An HDMI stick attached to a TV could bring the agent into your living room.
The “whatever else you’ve got lying on your workbench” line is the key. Meta is not prescribing a single form factor. It is saying that the code can run on whatever you happen to have around. That openness is the whole point of the project.
The Risk Warning
The “Proceed at your own risk!” warning is not trivial. Building with open-source code means you are working with code that has not been tested in every possible configuration. There could be bugs, compatibility issues, or unexpected behavior.
Meta is not hiding this. It is telling you up front that this is early-stage work. The company is making the code available, but it is not holding your hand through every step of the process.
What to Expect Next
The Muse Home Link waitlist is live now. Shipping is expected this month. The giveaway is a way for interested users to receive the device.
For anyone who wants a Muse device without building it from scratch, the waitlist is the path forward. For those who want to tinker, the open-source code is the starting point.
Our View
This is a smart move by Meta. Open-sourcing the code means anyone with a bit of technical skill can build a Muse device themselves. The projects Meta suggests are easy to understand and useful, and the fact that the company is giving away a physical device as well shows a genuine commitment to the idea.
The risk warning is fair. Open-source projects often require patience and troubleshooting, and Meta is not pretending otherwise. But the reward is worth it: a personalized AI agent running on your own hardware, built exactly how you want it.
This is a story about access. Meta is making its AI agent available to anyone who wants to build with it, and that is a change from the usual closed approach. The code is out there, the hardware is common, and the examples are clear. The only thing left is your own curiosity.
See the a run of 29 images at The Verge.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.