Muse Gadgets: Meta's Open-Source SDK for DIY AI Hardware

Muse Gadgets: Meta's Open-Source SDK for DIY AI Hardware

On October 2, Meta open-sourced the Muse Gadgets SDK, a toolkit that lets you wire Meta's Muse AI agent into your own hardware. Muse launched on September 6 as a personal AI assistant that books travel, fills out forms, and shops on your behalf. The Gadgets SDK gives it a body on your workbench: a screen, a speaker, a microphone, buttons, sensors, and actuators. The code lives at facebookincubator/muse-gadget-sdk on GitHub under Apache 2.0, with 2.1k stars and 423 forks as of this writing.

Meta's own words, from their Muse Gadgets page:

"Muse gadgets are open source devices you build yourself. 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."

"It's built by hackers, for hackers, just for fun. Side effects of tinkering may include bricked boards, voided warranties, brownouts, or bankrupties. Proceed at your own risk!"

Nat Friedman, head of product at Meta Superintelligence Labs, announced the release in a post on X: "Announcing Muse Gadgets, an open source ESP32 firmware and Linux sdk so that you can make hardware devices that work with Muse." He also revealed that Meta built its own gadget, the Muse Home Link, a USB-C-powered ESP32-C5 device with dual-band Wi-Fi 6, BLE, and 802.15.4 (Zigbee/Thread), 8 MB flash, and 8 MB PSRAM. Meta manufactured 5,000 units and is giving them away free to Muse subscribers in the US, limit one per subscriber, first come, first served. Meta's page says they "ship in October." A Discord community with 254 members online at the time of writing supports the project.

What the SDK actually does

The repository has two main SDK directories and a third for pre-built "skills":

ESP32 Device SDK. Meta describes it as: "Connect your ESP32 board to Muse through our open source SDK. Throw in a screen to show images, add audio in and out, or wire up other sensors." The supported chips in the article are the original ESP32, ESP32-C3, ESP32-C5, ESP32-C6, ESP32-C61, ESP32-S2, and ESP32-S3.

Linux Device SDK. "Turn that spare Raspberry Pi or Linux box into a Muse gadget. Hack in your own commands to let Muse handle sysadmin chores or your Home Assistant setup." Meta explicitly recommends the Raspberry Pi 5 for this path, but other SBCs may be used.

Skills directory. Pre-built integrations for Apple HomePod Mini, Brother and Epson printers, ESPHome devices, 3D printers, various TVs, and a range of smart-home gadgets. Meta's project ideas page also lists an M5Stack StickS3, an AiPi Lite desk buddy, an IDESpark ESP32 with a 1.9" color display, and a Home Assistant Voice Preview Edition. An HDMI stick for big-screen display is listed as "coming soon."

How pairing works

Before you flash or pair a gadget, grab an SDK token from the Muse Gadgets page and review the Gadget SDK Terms. Every gadget needs a token to pair. On the phone, open the Muse app (iOS or Android), go to Settings > Devices, turn on Developer mode, and look for devices prefixed with "MuseGadget."

What Meta recommends (and what we have)

Meta's own project ideas page lists the Raspberry Pi 5, the Waveshare ESP32-S3-Touch-AMOLED-1.75C (1.75" AMOLED, speaker, mic, battery), and the Seeed reTerminal E1002 (7.3" color e-ink) as featured builds. Meta adds a disclaimer: "Devices shown here are made and sold by other companies, and Meta doesn't endorse or warrant them." Neither the Waveshare nor the reTerminal is in the current ameriDroid catalog, but the CrowPanel line and our ESP32 dev boards cover the same chip and use-case space.

Raspberry Pi 5

Raspberry Pi 5

Meta's own words: "Give Muse a way to work with Home Assistant and other popular Linux applications." Quad-core Cortex-A76 at 2.4 GHz, VideoCore VII GPU, dual 4K HDMI, PCIe 2.0 x1, and a full USB 3.0 stack. Raspberry Pi 5 : from $134.95

ESP32-C6 SuperMini

ESP32-C6 SuperMini

The closest chip match to the ESP32-C5 inside Meta's free Muse Home Link. RISC-V core at 160 MHz, Wi-Fi 6, Bluetooth 5 LE with mesh, and 802.15.4 for Zigbee/Thread. 4 MB flash, USB-C, stamp-size footprint. Flash the ESP32 Device SDK and you have a sub-$10 Muse gadget that can join the same mesh as the Home Link. ESP32-C6 SuperMini : $6.95

CrowPanel 1.28 inch ESP32-S3 Rotary Display

CrowPanel 1.28" ESP32-S3 rotary display

ESP32-S3, 1.28-inch 240x240 round IPS touchscreen, and a rotary knob with push. The knob is a physical actuator out of the box, which maps directly to Meta's "buttons, sensors, actuators" interface. CrowPanel 1.28" : $29.95

CrowPanel 1.46 inch ESP32-S3 Rotary Display

CrowPanel 1.46" ESP32-S3 rotary display

ESP32-S3, 1.46-inch 360x360 round IPS touchscreen, rotary knob, and an RGB status ring. The ring is a ready-made "agent is listening / agent is speaking" indicator. CrowPanel 1.46" : $31.95

DEV-1 ESP32-C3 Development Board

DEV-1 ESP32-C3 (Apollo)

Designed and manufactured in the US. ESP32-C3-MINI-1, USB-C, GPIO, I2C, and an RGB LED. Pre-soldered headers. A no-frills ESP32 platform for a sensor or actuator node feeding data into Muse. DEV-1 ESP32-C3 : $12.95

Getting started

Clone the SDK from the GitHub repo. Each directory (esp32, linux, skills) has a README to get started and an AGENTS.md for coding agents. Pick the ESP32 path if you want a small always-on tile, or the Linux path if you want a full Home Assistant or sysadmin box. Flash or install, grab your SDK token, and pair from the Muse app. The skills directory is where you will spend most of your time if your hardware is not one of the pre-built targets: each skill is a thin adapter between the SDK's interface and your specific I/O.

If you are a US Muse subscriber, claim a Muse Home Link before the 5,000 units are gone. The ESP32-C6 SuperMini at $6.95 is a cheap way to build a second anchor or a peripheral sensor node on the same mesh.

Meta's SDK is provided "as-is, without warranty." The Apache 2.0 license does not cover the Jollybot avatar. Check your local laws before receiving RF signals or acting on Muse's instructions in a smart-home context.

Back to blog

Leave a comment

Please note, comments need to be approved before they are published.