Meta CEO Mark Zuckerberg closed the company's annual Connect event by revealing Muse Charm, a palm-sized wearable device built as a standalone hardware home for the company's Muse AI agent2,3. The device is expected to ship by the end of the year1.
Muse Charm resembles a chunky smartwatch without the strap, roughly the size of a keychain fob, with a ~2-inch OLED screen, front and rear cameras, 5G connectivity, and what appears to be at least three microphone holes on the front. A fingerprint sensor on the top right corner lets users activate the device and begin speaking to Muse without unlocking a phone or opening an app.
The Muse agent inside
The Muse agent is represented on the device's screen by a digital avatar the company internally calls "Jolly". Zuckerberg said the team packed "the whole Muse experience, including the whole real-time voice and avatar stack, into something that fits on a keychain and is always available to talk to". Users can carry the device in a pocket or clip it to a lanyard.
Zuckerberg described the reveal as a "one more thing" moment and acknowledged the hardware is not yet finalized; the team still needs to complete the component layout before a planned holiday-season shipment in December.
Entering a crowded, skeptical market
TechCrunch noted that the "AI amulet" category has become "a bit of a cliché in the tech industry," with devices like Friend turning the concept of physical AI into "little more than an exercise in eye-rolling for many consumers". ANALYSIS Meta's entry differs from earlier standalone AI pendants in one structural respect: Muse Charm is tethered to an existing, reportedly popular agent platform rather than launching a new AI service from scratch.
The Connect event also featured Meta's broader push into on-device AI processing. Earlier this week, Meta disclosed Private Processing, a confidential computing framework that runs AI models inside hardware-based trusted execution environments on its glasses line[3]. ◆ Muse Charm's 5G radio and dedicated hardware stack raise parallel questions about where inference runs and how user data is handled, though the Connect presentation covered those details only for the glasses product.
No pricing was disclosed during the keynote.