Realtime AI News
Tuya ships open-source skill letting Meta's Muse agent control 3,000+ device categories
Tuya Smart released Tuya Cloud Devices on November 12, an official open-source skill that lets Meta's AI agent Muse control smart-home devices already connected to the Tuya platform in natural language. The skill was originally built for the open-source agent OpenClaw and adapted to the Muse specification, with users pulling it from Tuya's GitHub repository and pairing it with a free API key.
Tuya Smart (NYSE: TUYA; HKEX: 2391) announced Tuya Cloud Devices on November 12, an official open-source skill for Meta's AI agent Muse that lets the agent control connected smart-home devices in natural language.
The release addresses a specific gap. Meta's Muse Gadgets is an open-source project that lets developers prototype add-on devices for Muse with off-the-shelf hardware, but the devices that stay in homes long term are mass-produced, always-online products backed by platform services. Tuya Cloud Devices is meant to bridge the two, routing Muse's instructions to hardware that is already deployed at scale.
Tuya says its platform covers more than 3,000 smart hardware categories across 200-plus countries and regions, with more than 2.09 million registered AI developers as of June 30, 2026. Once a device is connected to the Tuya AI Cloud, its thing model, cloud services and account system are already in place, so an agent can reach it through the official skill without any firmware changes.
There are two on-ramps. Muse users can get the skill from Tuya's official GitHub repository, apply for a free API key at tuya.ai, and hand both to Muse to control their Tuya devices by voice or text; Muse is currently available to users in the United States and Canada. Developers can use the open-source tuya-openclaw-skills repository to bring the same device capabilities into their own agents, or build AI hardware on TuyaOpen and the T5AI modules.
Portability is the design point. Tuya Cloud Devices was originally built for OpenClaw, an open-source agent, and was adapted to the Muse specification to enter Meta's ecosystem. As more agents adopt Markdown skills in the SKILL.md format, the same device capabilities can be reused across them, which means brands and solution providers no longer need a separate integration for every agent ecosystem.
Seen from the industry side, this is an infrastructure sample of personal AI agents moving from answering questions to getting things done. Beyond email and scheduling, home devices are the most natural next class of objects to operate, and the hard part was never whether a model understands a request to dim the lights but whether device-side semantics, accounts and cloud services are unified. Tuya is filling that layer with an existing IoT platform.
Tuya said it will keep extending the skill toward deeper integration across more device categories, camera visual understanding, energy-consumption analysis and scenario automation templates, with the aim of making one-sentence control a default capability of personal AI agents.
What to watch is practical availability. Muse is limited to the US and Canada for now, and the skill's real coverage, the reliability of device responses, and how data is handled in privacy-sensitive categories such as cameras will decide whether this moves from demo to daily use.
Why it matters
For Tuya, this plugs more than two million developers and 3,000-plus device categories into a new agent entry point; for Meta's Muse, controlling real household hardware is the step from chat assistant to something that executes.
Nearby Updates
All11/12, 18:56
Google widens Gemini 4 Carbon access with advanced coding features
Google has expanded access to Gemini 4 Carbon and added more advanced coding features to the model, according to a November 12 report from Jagonews24. The report does not disclose benchmarks, pricing or which regions are covered, so the feature list still needs to be checked against Google's own documentation.
11/12, 18:11
Anthropic begins IPO roadshow as report says its valuation has shrunk by $600 billion
Anthropic has begun its IPO roadshow, according to a November 12 report, taking the frontier model developer toward public markets. The same report states that the company's valuation has shrunk by $600 billion, framing the listing as a bet against an industry-wide pullback.
11/12, 16:47
London broadband ISP G.Network closes its first fully autonomous orders with an AI sales agent
London full-fibre ISP G.Network says an AI sales agent called "g" has completed its first end-to-end autonomous orders, handling product questions, plan recommendations, postcode checks and checkout without human involvement. The company estimates every £1 spent on AI usage costs returns £250 in sales, a vendor figure disclosed without methodology or sample size.
11/12, 15:55
VulnCheck questions Anthropic's Glasswing: 37,966 candidate flaws, 217 fixes on record
Security firm VulnCheck published a November 11 update arguing that Anthropic's Project Glasswing has produced far more vulnerability findings than fixes: 37,966 candidate discoveries by early November against 217 repairs in the ledger it tracks. Anthropic's own figures show 629 upstream patches and acknowledge that human verification, not discovery, is the bottleneck.