Choose what it hears.
System audio and microphone are separate. Keep your mic off when you only want the other side of the call.
LOCAL FIRST. CONVERSATION ALWAYS.
Meetings, interviews, a different language.
Keep up with the moment. Find the words that fit.
Your Mac. Your context. Your AI connection.
SAME SPACE. A DIFFERENT KIND OF HELP.
Three modes. One window.
Click a mode, or let it play.
FOR THE CONVERSATIONS YOU ALREADY HAVE
A CALMER WAY THROUGH THE CALL
You know the moment. A useful thought passes while you’re trying to write down the last one.
Flowtop keeps the thread close: what was said, the context that matters, and a little help with what comes next.
It all stays in one quiet window.
NO NEED TO START OVER
Connect Codex or Claude on your Mac. Or use a local model through Ollama. Your workspace stays the same.
Your provider’s plan and usage limits apply.
Connect your AITHOUGHTFUL IN THE SMALL THINGS
System audio and microphone are separate. Keep your mic off when you only want the other side of the call.
A short follow-up belongs to the conversation before it. Your selected notes and current transcript travel together.
Pin an answer. Ask for simpler words. Add detail when you need it. Help should match the way you think.
CLOSE TO HOME, BY DEFAULT
Speech is transcribed locally. Your workspace stays on this device.
AI requests send conversation text and selected context to the provider you choose.
Prefer local inference? Connect Ollama.
A FEW THINGS TO KNOW
Flowtop connects to your existing Codex or Claude CLI, or a local Ollama model. Your provider’s own subscription and limits still apply.
Only when you turn on the microphone. System audio and microphone have separate controls. Leave your mic off to capture the other side of a call.
The preview above is a live version of the interface with a sample conversation. Pause it and explore. It does not record audio, access your files or call an AI service.
The current build is for Apple silicon on macOS 14.4 or later. Audio capture permissions and a local speech model are needed for transcription.
Audio transcription and your workspace are local. AI requests include conversation text and selected reference context for your chosen provider. Account sign-in is separate from your AI connection.
READY WHEN YOU ARE
Version 0.5.1 · Apple silicon · macOS 14.4+
Personal development build. Not notarized yet.