Ski - A voice for your coding agent

sandra p b — Tue, 11 Aug 2026 09:42:09 +0000

You know the moment. You've handed Claude Code (or Codex, or Cursor, or whatever you're driving today) a real task , not "fix this typo," but "refactor the auth middleware and update the tests." It goes off and works. Even if you've turned off the "ask before edit" prompts and let it run end to end without interruptions, you still want to check in to see what it actually did, watch it reason through a tricky part, catch something before it goes further than you meant. That can take minutes. And in those minutes, you're either sitting there watching a terminal scroll, or you wander off and genuinely forget to come back and check.
Nobody was watching. That's the whole problem in one sentence: coding agents got capable enough to work unattended for real stretches of time, but the interface is still "stare at a terminal or forget about it."

So, what is SKI

SKI is a free desktop app that gives your coding agent a voice - in both directions.
You talk to Claude Code, Codex, Cursor, Gemini CLI, and a growing list of other agents by speaking instead of typing. And this is the part that actually changes the workflow - you hear them reply out loud. When the agent finishes a task or hits a decision point, it says so: "Tests pass, want me to ship it?" You don't have to be staring at the terminal to catch that. You can be in another window, another room, or heads-down on a different project entirely.
It lives as a small ambient widget on your desktop - a floating pill, or, if you're on a MacBook, a notch bar docked right under the camera housing , you hold a key to talk, and everything - speech-to-text and text-to-speech both runs locally on your machine.

Why "voice for coding agents" and not just "voice dictation"

This distinction actually matters, and it's the one thing I'd want you to walk away with if you read nothing else.
Dictation tools have existed for a while, and good ones too, they turn your speech into text and drop it wherever your cursor is. That's genuinely useful, and it's a one-way street: you talk, text appears, you're still the one reading the terminal and deciding what happens next.
SKI is built around a loop, not a transcript. The agent can also talk to you unprompted, when something needs your attention. That's a small architectural difference with a bigger behavioural consequence: instead of polling your terminal to see if something finished, you get told. Instead of typing a careful, well-formatted prompt, you can just say the messy version of what you mean, the way you'd explain it to a teammate leaning over your shoulder because you're not looking at a text box anymore, you're having a conversation.

Why this feels like it matters right now

A few things converged that make this a decent moment for something like this to exist:
• Agents got long-running. Multi-minute (sometimes multi-hour) autonomous tasks are normal now. The old "type prompt, watch cursor blink, read output" loop was built for instant responses, not for agents that go off and actually work.
• We talk noticeably faster than we type. For the messy, exploratory prompting that agentic coding rewards - "no wait, also check the edge case where the token's expired", voice removes a lot of the friction of typing-then-re-typing.
• Context switching is the real tax. If you're running Claude Code on one repo and Codex on an API server at the same time (which, increasingly, people do), the cost isn't the thinking , it's constantly checking both terminals. Having each project announce itself by voice -"From api-server: two tests fail, want the diff?" means you stop being the thing polling for status.

None of that requires SKI specifically. It's just the shape of the problem we thought was worth solving.

What it actually does, briefly

• Bidirectional voice loop - speak to the agent, hear it reply, with local barge-in so you can interrupt mid-sentence instead of waiting it out.
• Multi-project, multi-agent - run several projects at once, each with its own agent and its own voice, so you can tell by sound which one is talking to you.
• Works with most agents - Claude Code, Codex, Cursor, Windsurf, Gemini CLI, and roughly a dozen more through a shared skill file, not a hard integration per tool.
• Runs on-device - speech recognition and the neural voices are local. Nothing about the voice loop gets uploaded, and it works offline.
• Meeting recording, for free - record any meeting on your Mac, get an on-device transcript, and have your agent turn it into notes.

I'm not going to pretend it's magic. It's Mac and Windows only right now (no Linux yet), English only for now, and if you're in an open-plan office without a headset, voice coding is honestly just not for you yet — we'd rather say that plainly than pretend otherwise.

Also — it was Product of the Day on Product Hunt

A little extra context: SKI launched on Product Hunt and ended up #1 Product of the Day. That was genuinely nice to see, mostly because the people upvoting and commenting were exactly the audience this was built for - people who live in a terminal all day and immediately understood the "why didn't this already exist" reaction.

Try it, or don't but I'd like to know what you think either way

If you're someone who already spends your day driving Claude Code, Codex, or Cursor, I'd genuinely like your honest reaction to voice-in-the-loop including "this is a solution looking for a problem," if that's what you think. That kind of feedback is exactly why I'm posting this here instead of just writing ad copy for it.

• Website: heyski.io
• Product Hunt: producthunt.com/products/ski

Happy to answer questions about how the speech pipeline works, the on-device privacy model, or anything else in the comments.

DEV Community: sandra p b

Ski - A voice for your coding agent