Talk.
Walk away.
When a coding session finishes a turn, Ameka says two sentences in your ear — what happened, and what it wants to do next. You say "proceed" and it types your answer straight back into that session.
curl -fsSL https://ameka.ai/install.sh | bash
macOS 13+ or Windows 10+ · no API key · nothing leaves your machine · what it installs →
Four things happen. You do one of them.
No dashboard, no window to watch, nothing to check. The loop runs whether you are at the desk or three blocks away.
A session finishes a turn
Claude Code, Codex, a script, a cron job — anything that can run one shell command can hand Ameka a briefing.
You hear two sentences
"The webhook is wired and all the tests pass. Want me to promote staging?"
You answer out loud
"Proceed."
Ameka types it into that exact session
Straight into the right window in the Claude Code app — or silently into a tmux pane, on a machine that runs its sessions there. Then it waits for the next turn.
It works out what you meant.
There is no command list to memorise. A model reads what you said and decides what you were asking for — and anything it is unsure of gets read back to you before it is sent. Nothing is typed into a session without you saying yes.
Proceed.
Also: continue, do it, run it, ship it, keep going. Sends your go-ahead.
Hold on.
Also: no, stop, wait, cancel. Sends nothing at all.
Tell it to use Postgres instead.
Dictate anything. Read back to you first, then typed into the session.
Talking to Algo8.
Points her at another project and brings that session up on screen.
Open a session for Ractr.
Opens the folder in the Claude Code app, in the window you already have.
Why did it do that?
Also: details, more, explain. Reads you the longer version.
Nothing. In every sense.
No API key, no account, no bill
The voice, the listening and the two-sentence squeeze all run on your Mac — a neural voice model and whisper.cpp, installed for you. Bring an OpenAI key only if you would rather use their voice.
Nothing leaves the machine
The daemon binds to localhost and nothing is exposed off the Mac. No audio is stored, by us or by anyone else. There is no server on our end to send it to.
Two permissions, asked once
Microphone, so it can hear you. Accessibility, so it can type your answer back.
Then ameka doctor checks every engine and device, and names the one
command that fixes anything missing.
Ameka 01. A new computer.
Ameka Voice runs on any Mac and always will. It is also the first thing we built for the computer underneath it — new silicon, new memory, and an operating system conceived end-to-end by AI.