Write the commit message after the change lands.
/voice gave you the prompt. Now you need a sentence that explains the diff. Hotkey in the terminal, dictate, hit enter.
Voca is what you use everywhere else.
Anthropic shipped /voice for Claude Code in March 2026. It's good. Inside the Claude Code session, prompting by voice is now trivial. Outside it, the rest of your day is still typed: commit messages, PR descriptions, Slack threads with the engineer reviewing your branch, Linear tickets the agent's just unblocked, the doc you're updating because the feature shipped.
Voca is a system-wide voice layer that picks up where /voice stops. Same hotkey wherever your cursor sits. Works on Windows, macOS, and Linux.
No install for the demo, no card required.
/voice is the right tool for one job: turning your speech into the prompt you're sending Claude Code in this session. Push-to-talk, included in Pro/Max/Team/Enterprise plans, low friction.
What it doesn't do, by design:
/voice gave you the prompt. Now you need a sentence that explains the diff. Hotkey in the terminal, dictate, hit enter.
Walk through the change in plain English. Why, what, what's left. Voca drops the prose into GitHub's PR body the same way you'd dictate it into a doc.
The reviewer's left a question. Open the comment, hit the hotkey, answer the way you'd answer in person. Send.
Notion, Obsidian, README, internal wiki. Voca treats them like any other text field.
The Slack message that explains what shipped, who it affects, and what to look out for. Dictated in the time it takes to type the first sentence.
Other AI tools, other text boxes, other agents. Voca works in every one of them — including the apps that don't have a /voice equivalent of their own.
For everything that isn't a Claude Code prompt, you're back on the keyboard. That's the gap Voca fills.
A typical loop with Voca and /voice running together.
Same hand on the same modifier key for the entire loop. The keyboard's job becomes navigation and quick fixes. Everything that's prose is voice.
Describe the change you want, release. Claude Code makes the change.
Dictate the commit message, hit enter. Done.
Voca hotkey, dictate the PR description. Done.
Tell the team the PR is up. Done.
Hold spacebar, describe the next change.
Claude Code ships for macOS, Linux, and Windows (via WSL). A meaningful share of Claude Code users are on Linux because Claude Code is a CLI tool and Linux developers do most of their work in a terminal anyway.
The dictation tools that pair well with that workflow: Voca. The dictation tools that don't ship for Linux: Wispr Flow, Willow, Voibe, VoiceInk, Superwhisper, Monologue, Weesper, every other commercial option. If you're on Ubuntu or Fedora or Arch and you want voice for the parts of your day that aren't inside /voice, Voca is the option that exists.
Code in private repos shouldn't end up in someone else's model. Voca's posture matches Claude Code's own zero-data-retention guarantees: audio is deleted on transcription, never used for training, and metadata for billing — duration, timestamp, model — is the only thing we keep. Transcripts are stored only if you opt in.
Terminal, browser, IDE, Slack, Linear, your notes app, your git GUI.
Talk normally. Filler words and false starts get cleaned up.
0.4 seconds. Faster than you can switch tabs.
These aren't competitors. They're the two halves of a voice-first developer day. Also worth a look: Voca for Cursor, Voca vs Wispr Flow.
If you've rolled out Claude Code to your engineering team, the next bottleneck is everything that happens outside the CLI session. Voca's team plan covers that for $6/seat/month with a credit pool for light users, per-member usage caps, and one invoice on the 1st.
Free Voca tier gets you 30 minutes a month, no card. Plenty to test the loop on a real day of Claude Code work.
/voice works inside the Claude Code CLI only. Voca works in everything else on your machine, including your terminal, IDE, browser, Slack, and any text field. They complement each other.
Yes. They run independently and don't conflict. Most voice-first developer setups use both: /voice for prompting Claude Code, Voca for everything outside the CLI session.
Yes. Both Claude Code and Voca ship native Linux builds. Most other dictation tools don't ship for Linux at all. See dictation on Linux.
No. Audio is deleted as soon as transcription completes. We don't train models on your audio or your text. Transcripts are stored only if you opt in.
0.4 seconds median latency from end-of-speech to text appearing at the cursor. Built on the fastest open inference stack we could put in production.
Yes. 30 minutes of transcription a month, no card required. Enough to test it on a real day's work before paying.
$6/month flat or $0.10/hour pay-as-you-go. Same hotkey on Mac, Windows, and Linux.