Aftertone
A free hook that speaks a short summary out loud after your coding agent finishes a reply, using a local ONNX text-to-speech model with no cloud API, API key, or subscription.
Aftertone runs a local text-to-speech daemon (Supertonic via ONNX Runtime) and registers a hook with a coding agent so that when a reply finishes, a short spoken summary plays. The same tts_daemon backs adapters for Cursor and Claude Code today, with Codex and OpenCode adapters in progress.
Installation is a shell or PowerShell script that installs to a local directory, downloads the models, registers the hooks, and starts the daemon. Speech is opt-in per chat session through a slash command such as /aftertone_on in Claude Code, and agents are expected to end substantive replies with a specific tag; without the tag, nothing is spoken. Toggling and status commands work per session so other sessions stay silent until enabled.
It is MIT licensed and free, with no API keys, subscription, or cloud TTS service involved.