What is the best dictation tool for Claude Code?

Use the built-in /voice first. It is already in Claude Code, it understands common developer terms, and it does not need a second app or a second account. Switch to a system-wide dictation tool when the built-in option cannot run in your setup, or when the processing it uses does not fit your work. The rest of this page explains the trade-off, so you can check it against your own setup. Our guide to voice input for Claude Code and AI coding tools covers the other coding tools in the same way.

The built-in /voice is the default choice

Claude Code's dictation is turned on with /voice, and you speak into the prompt you are already typing. Its requirements are narrow. You need a claude.ai sign-in, and the feature is not available with an API key or a cloud provider ¹. You also need a local microphone, so it does not work over SSH or in cloud sessions ².

Its main strength is that it is tuned for the words you type while coding. The docs say transcription is tuned for coding vocabulary in both hold and tap modes, and they give regex, OAuth, JSON and localhost as examples of terms it recognizes ³. Those are the vendor's own claims, and the page does not publish an accuracy figure, so check the transcript before you press Enter.

It also does not use up your Claude allowance. Transcription does not use Claude messages or tokens, and it does not count toward the limits in the usage view ⁴. The reply to your prompt still counts as usual.

The one real cost is privacy. The audio is not processed on your computer. Claude Code streams it to Anthropic's servers for transcription ⁵. For many people that is an acceptable trade for a tool that works directly in the prompt. For confidential code or regulated material, it may not be.

When a system-wide app is the better choice

A system-wide dictation app types into whichever text field has focus. That makes it the fallback for the cases where /voice does not run, such as an API key setting, a remote session, or a terminal with no microphone permission. Our troubleshooting page for Claude Code voice covers how to fix those first, so you can decide whether it is worth the switch.

The choice inside this group turns on processing. Superwhisper lists a free plan, and its documentation says that on that plan, dictation uses local Whisper models without AI processing, and the company says nothing dictated leaves the device ⁶. That fits a privacy-first workflow, though you should check the current plan terms before you rely on it.

Wispr Flow takes the cloud route. Its privacy page says dictated audio goes to the company's servers over an encrypted connection, so cloud processing is part of its design ⁷. It is a fit where you accept cloud processing in exchange for its features, and it is not a fit where the audio must stay on the device.

Yap, which we make, is a dictation app for Mac and Windows that runs on your device. It is one of the options above, not the only one, and you should test any of them in your own terminal before you depend on it.

Test the terminal before you trust any tool

Terminals are not ordinary text boxes, and a system-wide tool can insert text in a different way from the one you expect. Wispr Flow's help article for coding tools says that on a Mac, in Claude Code and Codex CLIs, it pastes long dictations in parts ⁸. It also says that in supported coding editors on Windows, it uses Shift+Insert instead of Ctrl+V ⁹. Those are the vendor's own descriptions.

The test is short. Place the cursor at the Claude Code prompt, dictate a two-sentence prompt that includes a file name and a function name, and read what appears before you press Enter. If the file name is wrong, add it to the tool's custom vocabulary if it has one. A custom vocabulary is a list of terms the recognizer is told to expect, and it helps most with names your team uses every day. For the mechanism behind text insertion, see our explanation of system-wide dictation.

What to do

1. Start with /voice if you have a claude.ai sign-in, a local microphone and no problem with cloud processing. Turn it on with /voice, then hold Space to speak.
2. If /voice cannot run, fix the sign-in or permission problem first, using the Claude Code voice troubleshooting page.
3. If the audio must stay on your computer, choose a system-wide app that states on-device processing, and confirm the claim in its own settings.
4. In every case, read the transcript before you send the prompt, and correct file names and identifiers by hand.

For the wider picture of where to speak and where to type, see our guide to dictation in your apps, and the question of whether Windows voice typing works in every app for the same tests on another platform.