Meta now lets developers run its Muse AI agent on their own hardware. On October 2, the company open sourced the SDK (software development kit) and firmware for Muse gadgets, with support for ESP32 (an off-the-shelf microcontroller board) and Raspberry Pi. The code is published at github.com/facebookincubator/muse-gadget-sdk under the Apache License 2.0.
What can you build with the Muse gadget SDK?
Meta suggests three starter projects for Muse gadgets: a color E Ink display showing AI-powered reminders, an HDMI stick to display Muse on a large screen, and a small touchscreen resembling a DIY version of the Muse Charm. “Program an off-the-shelf ESP32 board or set up a Raspberry Pi with our SDKs, then connect Muse to your displays, buttons, sensors, actuators, and whatever else you’ve got lying on your workbench,” Meta writes in the repository.
The warning is direct: “Side effects of tinkering may include bricked boards, voided warranties, brownouts, or bankruptcies. Proceed at your own risk!”
How does pairing work?
Every Muse gadget needs an SDK token before it can pair with the Muse app. ESP32 and Linux gadgets connect via Settings > Devices on iOS or Android after enabling Developer mode; devices appear prefixed with “MuseGadget.” Each project directory in the repository includes a README.md for human setup and a separate AGENTS.md file formatted for coding agents such as Muse Code.
What is the Muse Home Link?
Alongside the open SDK, Meta is distributing a ready-made gadget it built itself, called the Muse Home Link. Using community-built skills, it can “turn on a light, control your TV, send a document to a printer, and more, depending on your setup,” according to Meta. Meta manufactured 5,000 units, according to Nat Friedman of Meta Superintelligence Labs, per The Verge. A waitlist is open now; shipping is planned for sometime this month.
What this means for AI-search visibility
Here is the simple version: whoever builds the skills that run on Muse devices controls the answers those devices give. That is the new distribution question for brands.
When someone asks a Muse Home Link to find a service or control something in their home, the agent fulfills that request through community-built skills. Think of skills the way you’d think of apps on a phone. The brands and services that get integrated into those skills early are the ones that get surfaced. In Hingewise’s view, those that wait may find it harder to get in later.
The AGENTS.md pattern in the SDK is worth a separate look. Each project directory includes a file formatted specifically for AI coding agents like Muse Code to read and act on, not just human developers. It is an early but concrete sign that AI agent-to-agent coordination (one AI helping configure another AI’s interface) is moving from research into shipping developer tooling.
Meta’s approach here follows a familiar platform logic: open the hardware layer, keep the AI agent itself in Meta’s hands, in Hingewise’s read. This expands Muse’s physical footprint without Meta bearing the manufacturing cost at scale. The 5,000-unit Home Link giveaway is small, but it is likely aimed at seeding the community and produce early real-world integrations that others then build on.
In Hingewise’s view, the home AI agent layer is the next surface after AI Overviews (Google’s AI-generated answer panels in search results) and chat-based AI. Today, visibility strategy focuses on which brands get cited in a ChatGPT response or a Perplexity answer. The question forming now is: which brands get integrated into the skills a Muse agent calls when someone makes a request at home? Community-built skills are the answer, and the window to participate early is open right now.
What to watch: the pace of community skill development in Meta’s Discord, which categories of home integration get built first, and whether Meta opens more of the Muse agent’s action layer to third-party services beyond the current hardware SDK.
Sources: The Verge, October 2, 2026; github.com/facebookincubator/muse-gadget-sdk
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.