Talk to your
AI coding agent.

Voice commands for Claude Code, Codex, and any CLI tool. 100% local — no cloud, no subscriptions.

$ brew install dragfly/tap/dictare
$ curl -fsSL https://raw.githubusercontent.com/dragfly/dictare/main/install.sh | bash
Permissions required — macOS and Linux need a few permissions to work. See setup guide →

macOS & Linux · MIT License · View on GitHub →

For the best experience, watch on YouTube in 1080p

The key difference

No focus required.

Dictare listens globally. Speak while reading docs in your browser, reviewing a PR, or doing anything else — no need to click back to a terminal. Your agent hears you wherever your focus is.

100% Local

Whisper, Parakeet — your choice of speech model. Zero data leaves your machine. No accounts, no API keys, no cloud.

Always Ready

Runs as a system service. Models preloaded at startup — speak and it responds instantly. Zero cold-start.

One Interface, Every Agent

Claude Code, Codex, Gemini CLI, Aider, Pi — same hotkey, same voice commands. Switch agents mid-session with your voice.

1

Install as a service

Run dictare service install — Dictare starts at login, preloads STT models, and waits silently in the background.

2

Launch your agent

Run dictare agent my-project and start speaking. Claude Code — and all your favourite AI coding agents — executes. That's it.

3

Switch agents on the fly

Say "agent personal" or "agent work" mid-session to switch to any configured tool.
Voice triggers are multilingual — EN, DE, ES, IT, FR out of the box.

Claude Code Codex Gemini CLI Aider Pi Any CLI tool

Start talking to your code.

Open source, free forever. Built for developers who think faster than they type.