Tama

Tama turns talking into notes. Say what's on your mind out loud, and it lands as a plain markdown file in a folder on your own machine a few seconds later. No app to open first, no account, no subscription.

That folder is the start of what people call a second brain: one place where everything you know lives, in plain files, linked together, that an AI can read and write to instead of you re-explaining your life to a chatbot every time you open one.

Most notes never get that far. A Notion page you never reopen, a notes app with 400 untitled entries, screenshots you saved and forgot; that's where ideas usually go to die. A second brain works because it's the opposite: not scattered apps, one folder a model can read in full and actually use. The full argument for why that matters is here: A Second Brain That Talks Back.

A small device you can talk to from anywhere in the house, paired with software that runs entirely on hardware you own. The bet: capturing a thought should be as fast as talking, and your notes should never have to leave your machine to make that true.

• Now: talk into a phone or a terminal and it lands in your second brain. Ask a question, get an answer back out of your own notes.
• Next: a web app to chat with it, and the device you talk to without picking anything up.

You talk, tama-server turns it into text on your own machine, and it lands as a dated markdown note in your second brain. Nothing leaves the machine, and there is no language model in that path, so there is no account and no bill.

Your second brain is a plain folder, so anything can read it. An MCP layer is coming that hands it to Claude Code, Cursor or any MCP client directly, with no copy and no upload.

Apache-2.0, all repos.