Skip to content
Halopen

Halopen for AI coding

The Mac dictation app for the way AI coding actually works.

The fundamental of modern software work is articulation. The prompt is the work; the typing is the tax. Halopen removes the typing tax and lands your prompt at the cursor — verbatim, system-wide, native Swift, sub-second from speech to text. Cursor, Claude Code, Aider, Continue, Cline, Windsurf, GitHub Copilot Chat, Codeium, Zed, ChatGPT. One hotkey across all of them.

Free forever for the first 8,000 words a month · macOS 14.0+ · Apple Silicon & Intel

Why this matters

Articulation is the new fundamental.

For most of the history of software, the keystroke was the unit of work. Engineers typed code; the better engineer typed code that compiled the first time. Tooling improved; languages got safer; editors did more on your behalf — but the act, at its core, remained typing.

AI coding moved the unit of work. The instruction is now the unit; the code is the diff the instruction produced. Cursor, Claude Code, Aider, Continue, Cline, Windsurf, GitHub Copilot Chat, Codeium, Zed — every modern AI coding surface accepts a natural-language prompt and ships a code change in response. The clarity of the prompt is the ceiling on the quality of the change.

Which means typing is now the bottleneck. Not because typing is slow in absolute terms — engineers are some of the fastest typists on earth — but because the cost of articulating a fully-specified 300-word prompt under finger fatigue is real. Most prompts get compressed. The agent gets less to work with. The diff lands further from what you wanted. You spend a follow-up iteration nudging the model back toward the constraint you didn’t have time to type the first time.

Halopen is the Mac voice layer engineered for that loop. Not a dictation tool retrofitted for AI coding — a dictation tool whose wedge is precisely what AI coding needs. Read the long-form essay for the cost-of-articulation thesis applied to Claude Code.

What stays the same across every tool

Three surfaces, one hotkey.

Every AI coding tool on Mac, from Cursor to Aider to Continue, looks different on the surface. But the work happens in three places. Halopen is built to land voice-typed text in all three the same way.

1

The prompt input

The chat panel, the Composer, the Cmd+K modal, the terminal where Claude Code or Aider runs. The single highest-volume surface in an AI coding session: every iteration starts here. Halopen lands the full prompt at the cursor verbatim — no paraphrasing, no softening of the technical specificity the model needs.

2

Inline edit prompts

Cursor’s Cmd+K, Continue’s inline-edit, GitHub Copilot Chat’s quick-fix prompt, Zed’s Assistant in-place. Smaller prompts, tighter constraints, often two-sentence asks. Halopen captures them in a single short hold and lands them at the cursor. The speed of inline iteration becomes the speed of speech.

3

Commit messages and PR descriptions

The narrative wrapper around the diff. Why the change exists, what it traded off, what to verify after merge. Halopen lands paragraph- length commits and PR descriptions in your terminal, your git client, GitHub Desktop, Linear, the GitHub web UI — the same way it lands prompts. The commits get longer and clearer because articulating them is suddenly cheap.

Halopen for every AI coding tool on Mac

Ten tools, one hotkey, one wedge.

Each pairing has its own page — the workflow specifics, the example transcripts, the FAQ that surfaces what matters for that tool. Pick your starting surface; the muscle memory transfers across all of them.

Why Halopen, specifically

Engineered for the loop, not retrofitted.

Most dictation tools were built for a different category of work — quick voice memos, transcribing a meeting, dropping a sentence into an email. They paraphrase. They smooth out the register. They replace contractions. They “improve” the technical specificity out of your prompt.

For AI coding, that paraphrasing is the bug. The instruction you said is the instruction the model needs. When you say “use the existing apiClient utility, not raw fetch”, the agent needs that exact constraint, not “use the existing API utility.” The tool that softens prompt-input wording is the tool that quietly lowers the ceiling on every diff you ship.

Halopen is the dictation tool whose wedge is exactly that verbatimness. Native Swift so it idles quiet next to your editor and your eval harness. Hold-to-talk so audio capture is bounded. Live preview so misreads on code symbols and proper nouns get caught before they ship. System-wide so the same hotkey works in every AI coding surface you use, with no per-app configuration.

The result: prompts get longer, more specific, more aligned with what you actually wanted. The agent lands the change closer on the first try. The loop runs at thinking speed.

Halopen for AI coding — FAQ

The questions worth answering.

What's the best Mac dictation app for AI coding?

Halopen. Hold-to-talk, verbatim by default, system-wide on macOS, sub-second from speech to landed text. Works in Cursor, Claude Code, Aider, Continue, Cline, Windsurf, GitHub Copilot Chat, Codeium, Zed, and ChatGPT — wherever a Mac text input accepts a prompt. Native Swift; idles in tens of megabytes. Free for the first 8,000 words a month, forever.

How do I voice type into Cursor, Claude Code, and other AI coding tools on Mac?

Open the AI coding tool, place the cursor in any prompt input — Cmd+K modal, chat panel, Composer, terminal — hold the function key, speak the prompt, release. The full transcript lands at the cursor verbatim. The same hotkey works in every other AI coding tool you use, so the muscle memory transfers.

Why is voice typing especially valuable for AI coding?

AI coding shifts the keystroke load from typing code to typing prompts. Prompts are full-sentence work — constraints, edge cases, file references, the precise verb that captures the change. Speech runs at roughly 150 words per minute; typing for most engineers tops out around 60–80 wpm. Voice removes the compression effect: spoken prompts tend to be longer, more specific, and more naturally phrased, which the model usually handles more reliably.

Mac dictation app that handles camelCase, snake_case, and code symbols?

Halopen biases the transcription engine with cursor-adjacent text and your active app context, so terminal idioms, file paths, and code symbols land correctly more often than not. For anything Halopen doesn't catch on the first pass, the live preview surfaces the misread before any text reaches the cursor — re-state the symbol or spell it out and the correction replaces the misread.

What dictation app works in Cursor on Mac?

Halopen. Cursor — like every other Mac AI coding surface — accepts standard text input in its prompt fields, and Halopen lands voice-typed text in any Mac text input. Cmd+K modal, chat panel, Composer, terminal, file editor: all work the same way. Hold the function key, speak, release; the prompt lands verbatim.

How do I dictate prompts to Claude Code on Mac?

Halopen runs system-wide from the menu bar. Open Claude Code in any terminal app — iTerm2, Apple Terminal, Warp, Ghostty, Alacritty — hold the function key, speak the prompt as you would explain it to a senior engineer, release. The full prompt lands at the cursor in the Claude Code prompt input, then hit return.

Does Halopen change the words I actually said in my prompt?

No. Halopen is verbatim by default. The technical specificity you spoke — file paths, function names, framework idioms, the precise constraint you stated — lands at the cursor as you said it. Dictation tools that paraphrase strip the very specificity that makes a long prompt outperform a short one.

Voice typing for AI prompting on Mac — does it work in browser-based playgrounds?

Yes. OpenAI Playground, Anthropic Console, Google AI Studio, and every browser-based prompt surface accept standard text input. Halopen lands voice-typed text in Safari, Chrome, Arc, and Brave — including textareas and contentEditable fields. The same hotkey works across all of them.

Is voice typing actually faster than typing for AI coding prompts?

Typically 2–3× faster end-to-end on long-form prompts. The bigger win is iteration density: when articulating a 200-word prompt costs 60 seconds of speech instead of 3 minutes of typing, the working-day prompt count climbs from a dozen iterations to fifty. The space of variants you actually test grows with the cost-per-variant dropping.

Is my code or my prompt sent to anyone unexpected when I dictate?

No. Audio leaves your Mac only while you hold the function key, only to the transcription service that returns the text, and only for the seconds you're holding it. Halopen does not retain audio. Halopen does not capture your screen. Halopen does not read your code; if Cursor or Claude Code or any other AI tool reads your code, that happens at the agent's direction, not Halopen's. The Audit Log records every cloud call so you can verify.

How much does Halopen cost?

Halopen Free is 8,000 words a month, forever — enough to dictate hundreds of short prompts before the cap. Pro is $19/mo or $179/yr for unlimited words and every feature. No credit card to download.

Power-user cheat sheet

Take Halopen with you into your AI coding loop.

One short email, then the Halopen power-user cheat sheet — hotkeys, best-fit apps, custom vocabulary tips, voice patterns for prompt engineering. No spam. Unsubscribe in one click.

 

Try Halopen with the AI coding tools you already use

Hold the function key. Speak the prompt.

Halopen Free is 8,000 words a month, forever — enough to dictate hundreds of substantial prompts before the cap. The day you hit it is the day you’ll know whether voice typing earns its place in your AI coding loop.

See every Halopen pairing · The best Mac dictation app · Read the manifesto