Author: P.J. Leimgruber
OpenHome Now Controls Your Computer
Until now, OpenHome has lived on the edge, or in the cloud (sounds blissful) — processing your voice, routing your requests, connecting to services. That's still true. But today, we're opening up a new frontier: your local machine.
With two new developer templates in our abilities library, you can build voice-controlled abilities that reach directly into your computer. Say "check disk space," "start my dev environment," or "run a git status," and OpenHome will do it. Not through a third-party integration. Not through a cloud relay. On your actual machine, in real time.
This is a new category of what OpenHome can do.
OpenHome is the Voice Layer OpenClaw Has Been Missing
If you're already building with OpenClaw, you know the power of a locally-running AI agent that can actually do things on your machine. What OpenClaw doesn't have out of the box is a polished, always-listening voice interface — a way to talk to your agent naturally, from across the room, without typing a command.
That's exactly what OpenHome provides.
Pair OpenClaw with OpenHome and your local agent gains a production-quality voice layer: wake word detection, streaming STT, natural TTS, conversation memory, and a growing marketplace of abilities to extend what it can do. Your OpenClaw automation workflows become things you can just talk to. The agent you've been building in the terminal becomes something that lives in your space.
For OpenClaw developers, this is the missing piece. You've already done the hard work of building intelligent local automation. OpenHome turns it into something you can actually use hands-free.
Two Templates, Two Paths
We've shipped two templates to get you started — one for simplicity, one for power.
Local Link (Mac)
Local Link is the fastest path to local control. Setup is a single Python script, your OpenHome API key, and a running terminal. That's it.
You speak a command in natural language. OpenHome converts it to a terminal command, sends it to your Mac via the local client, runs it, and speaks back the result in plain English. No other software required, no cloud dependencies for execution.
It's ideal if you want quick access to Mac workflows — file management, system info, development scripts, application launching — without spinning up a complex local stack.
[View the Local Link template on GitHub →]
OpenClaw Template (Windows, macOS, Linux)
For developers who need more — cross-platform support, LLM-powered intelligence on the local side, and robust automation — we've built a template on top of OpenClaw, the open-source local AI agent.
OpenClaw brings its own LLM reasoning to the table, meaning your voice commands don't just execute as terminal strings — they're processed intelligently at the local level too. It's more setup (Node.js, a daemon, an LLM API key), but it unlocks more sophisticated multi-step workflows and works across all major operating systems.
[View the OpenClaw template on GitHub →]
What You Can Build
Both templates are starting points, not finished products. The real power is in what you customize them into. A few directions developers are already exploring:
Dev environment controller — one voice command opens your IDE, starts your local server, and opens the browser. "Start coding session" and you're in.
Git assistant — check status, commit changes, push branches, create new ones. No switching windows.
System monitor — CPU, memory, disk, battery, all read out on demand in plain language.
File and folder automation — organize downloads, back up directories, find large files.
Morning routine launcher — open Calendar, Slack, and your email in a single command.
The templates include safety primitives — command validation, confirmation prompts for destructive operations, timeout handling — so you're not building on a blank canvas.
How It Works
Both templates use the same exec_local_command() function in the OpenHome SDK. Your ability captures the voice input, sends it to the local client running on your machine, gets back a result, and uses an LLM to convert that output into a natural spoken response. The whole loop happens fast enough to feel conversational.
The key difference between the two templates is what's running on your machine: Local Link uses a minimal Python script that executes terminal commands directly; OpenClaw runs a full local daemon with its own LLM layer for more intelligent execution.
Get Started
Both templates are live in the OpenHome abilities library and on GitHub. Clone the repo, copy the template, customize the trigger words and logic for your use case, and you're building.
If you ship something interesting, we want to see it. Share it with the community, open a PR, or drop it in the discussions tab — exceptional community abilities get promoted to official status and a permanent credit.
Update on DevKits
We've been overwhelmed by the response from developers wanting to build on OpenHome hardware — and we're grateful for every application. Because demand is high and DevKits are limited, we're prioritizing developers who are active in our community: folks opening issues, submitting abilities, participating in discussions, and sharing what they're building.
If that's you — or if you're ready to become that — we'd love to have you.
The DevKit gives you direct access to the OpenHome hardware platform: a purpose-built voice device with hi-fi audio, a beamforming microphone array, WiFi 6, Bluetooth 5.0, Matter, Thread, and the full OpenHome SDK to build against. It's the fastest way to go from idea to working voice ability on real hardware.
Apply for a DevKit at dev.openhome.com →
The best thing you can do to move up the list: get active on GitHub. Star the repo, open a discussion, submit a template, or share something you've built. We read everything and it genuinely influences who gets hardware first.