Your meetings and notes
don't deserve to be
on the internet.
Cocoon is one step toward a future where AI is powerful and private — not one or the other.
Cocoon started as a personal frustration. Every meeting app I tried sent audio to someone's server. Every AI note-taker required an account, a subscription, and implicit trust that a company wouldn't misuse what I said.
I believe that changes in your meeting with your team, the idea you captured in a voice memo at midnight, the conversation with your doctor — none of it should pass through someone else's infrastructure. Not even for a fraction of a second.
So I built Cocoon. A single-person passion project. No VC funding, no growth team, no ad revenue to chase. Just an app that does exactly what it says: AI that lives entirely on your device.
You don't need a server. You don't need an ad. You just need your phone and 5GB of storage.
Built on three promises
No servers. Ever.
Cocoon has no backend infrastructure. There is no server receiving your audio, no database storing your words, no engineer who could ever access your meetings.
No ads. No tracking.
No advertising SDKs. No analytics. No crash reporting services. No attribution frameworks. Nothing listening. Nothing watching.
AI that respects you.
Whisper and LLaMA run entirely on your device via llama.cpp. The intelligence is real — it just doesn't need to phone home to work.
One step toward privacy-first AI
The AI industry has normalised sending your most private moments — your voice, your ideas, your decisions — to distant servers to be processed and stored. Cocoon is proof that this trade-off is unnecessary. Whisper and LLaMA run on your iPhone. The intelligence is real. The privacy is absolute. This is what AI should look like.
Ready to take your privacy back?
© 2026 Cocoon · Made with care for your privacy.