Direct your AI coding agents by voice, without looking at the panel. TUNLstack is the voice command deck for Claude Code and Codex. Every reply reads itself aloud, word by word, you answer by voice, and the Cockpit shows every session's status at a glance. It is a Windows desktop app that runs alongside your coding agent, in VS Code or in a terminal, and your agent stays exactly as it is.
Almost certainly. There is no minimum-spec wall: TUNLstack adapts to the hardware it finds and never refuses to run. The whole app holds about 1.5 GB of RAM and needs no graphics card. Optional chat summaries are a separate download of up to about 7 GB and use more memory while they run. On machines under 12 GB of RAM (or with 2 performance cores or fewer) it runs in a lighter mode that keeps the offline dictation model unloaded; dictation still works.
For the full multi-session workflow, we recommend Windows 11, 16 GB of RAM, and 8 logical cores, with 32 GB if you keep many parallel windows open. The extra headroom is for your editor and your agent sessions, not for TUNLstack: each additional VS Code window with a coding-agent session inside costs roughly 1 GB.
Claude Code and Codex in VS Code, and Claude Code, Codex, OpenCode, Pi, GitHub Copilot CLI and Gemini CLI in a terminal. Claude Code works out of the box; the others connect during setup.
No. TUNLstack injects zero tokens into the model context: it reads what Claude already said and adds nothing to the conversation. Your bill and your context window are untouched.
Founders Edition: buy once, yours to keep. $49, one time. No recurring fee, no tiers.
Eight days, the whole app, nothing gated, no card required. Download, and it starts reading. A creator or community code adds three more days, eleven total.
No. The voice readout runs locally on your PC; no cloud voice account, no per-word meter. Eleven voices, each with its own pace setting.
TUNLstack never routes anything to a model you didn't pick: it reads what your agent already said. Its optional chat summaries are written on your PC, by a model you choose to download. The voice readout is synthesized locally on your PC. Dictation uses a fast cloud speech-to-text service by default.
No. Claude Code stays exactly as it is; TUNLstack runs alongside it, in VS Code or in a terminal. Download, and it starts reading. Three keys to learn: talk to it, steer it, move through it.
The trial is the answer to that: eight days with nothing held back, no card required. And if you buy and change your mind, see the refund policy at /refunds.