Product

Your people, devices, and the AI model you choose

Koan builds a map of your people from conversations and contacts, a map of your devices and apps — and answers using the model you pick. Adjustable autonomy, all on your computer.

The current beta is a clear scenario: connect Telegram and your phone, ask about people, and give tasks to devices.

One window for email and other messengers is planned after a successful beta.

What you get
  • Answers from real messages, not invented text.
  • Tasks from your phone executed on your computer — with a detailed report.
  • Autonomy you control: from confirmation to automation.

Connect Telegram and your phone

Step-by-step setup with hints. Koan builds a map of people and a map of devices.

Ask or give a task

Koan finds the relevant messages or the right window — and answers or executes to the point.

Set your autonomy level

Koan starts by showing drafts and asking. As trust grows — routine runs itself.

Read-only email (optional)

You can add a mailbox: Koan reads the messages but sends nothing.

Beta scenarios

Scenarios we are testing now

Beta participants run them on their own — on their Telegram and their model, without a developer on the call.

01

“Did the money from Mark arrive?” — Koan answers from your Telegram chats or prepares a draft reply.

02

“Gather analytics on my computer” — Koan runs the task on your computer from your phone and sends the report.

Why Koan

It knows your context — even with a small model on your computer

Koan pulls together everything it knows about a person and your devices into a single context, and gives the model exactly what it needs. You do not retell anything, and you set the autonomy level.

What stays under your control
  • Which model to use — local on your computer or in the cloud.
  • The autonomy level — from confirming every action to full automation of routine.
  • The phone app (Satellite) and the cloud — optional, not required.

Next step

Try Koan in the next beta wave

Apply if you have Windows, live Telegram, and a local AI model (for example via Ollama). You will help us test answers from chats and read-only email.