— Open source
The engine is open.
Everything that hears you, transcribes you, and types for you — six Rust crates and
the rekody CLI — is on GitHub under MIT.
The free Mac app is a closed-source shell on the same engine, so the code that touches
your voice is code you can read.
— The workspace
Six crates. One binary.
Each crate has one job: microphone capture and voice detection, on-device Whisper transcription (offline, out of the box), optional bring-your-own-key cloud engines, the global push-to-talk hotkey, cursor-level text injection, and the orchestrator that ties them together. TOML config. No telemetry.
The license is MIT — nothing fancy. Use it at work, fork it, bundle it with a closed-source product; the only requirement is keeping the copyright notice.
Rather than us describing the code, read it — the workspace is small enough for an afternoon.
— Install the CLI
One command.
No installer, no account. The binary is ~35 MB, and the on-device Whisper model (~75 MB) downloads once at setup — after that, dictation runs offline.
Prefer building from source, cargo, or the curl installer? Every path is in the README.
macOS 13+ · Apple Silicon & Intel · Accessibility + Microphone permissions
— Contribute
Issues, PRs, and opinions welcome.
The repo is small and so is the team. Bug reports with what you ran and what happened help most, and tight diffs are the fastest way to a merge. Start with an issue, read CONTRIBUTING, or bring design questions to Discussions.