Launched this week
Tamekoshi
A Tamagotchi-style AI companion that does real work
76 followers
A Tamagotchi-style AI companion that does real work
76 followers
Tame is an AI companion that does real work from the top of your screen: one message with three asks becomes three subagents in parallel, it runs tasks on its own computer, answers clients on WhatsApp and Telegram and writes meeting notes speaker by speaker. Nothing acts without your OK. Think Tamagotchi, all grown up. Free beta; a mobile app, multiplayer and open source are on the roadmap.












Free
Launch Team / Built With

Memanto CloudYour agents focus. Memanto remembers.
Promoted


Hey Product Hunt π
Big week on the internet. A certain open-world game finally came out yesterday, 13 years after the last one, with no PC version at launch.
So we're launching too: no 13-year wait, and PC on day one. π«‘
Why a Tamagotchi?
Remember the little egg on your keychain that beeped until you fed it, and quietly died during math class? Tame is that idea, all grown up. It still lives on your screen (right at the top), still has moods, still lights up when you show up. The difference: you don't take care of it. It takes care of your to-do list.
What it does
Ask for three things in one message: it splits them into three subagents that work side by side, and each one reports back on its own
Runs real tasks on its own computer (browser, terminal, even an Android phone) and shows every step
Nothing acts without your OK: anything that writes, sends or spends waits for your approval
Answers your clients on WhatsApp and Telegram, and you can take over any time
Meeting notes with speaker labels, transcribed on your computer, with no bot joining the call
Pages: a chat becomes a small live app (orders, stock, margins)
Rooms where companions work together, GIFs included
286 apps to connect, 80+ AI providers behind one picker, 40+ messaging channels, 20 languages
Friendly by design
Lots of AI agents are racing to be the edgiest one in the room. Tame would rather be the one you're happy to see in the morning: friendly, a little nostalgic, and careful with your stuff. It asks before it acts, tells you what it did, and celebrates a little when the work is done. π₯
What's next
π± A mobile app: Tame back in your pocket, where Tamagotchis belong
π₯ Multiplayer: your companions and your team's, working in the same room
π Open source: we're getting the code ready to open
It's a free beta, so expect rough edges. Tell us what breaks, what you'd want Tame to do next, and whether you had a Tamagotchi (and what happened to it πͺ¦).
Daniel
Will the mobile app work without the desktop app running?
@relli7Β Hey David, yes, it will. There's still a lot on our list, and a cloud version is one of the big ones. Once that's out, you'll be able to connect the mobile app straight to your own instance, no desktop needed.
does it get annoying having it sit at the top of the screen all day or can u tuck it away
@igorblinkΒ Not annoying at all, it stays out of the way. By default Tame is tucked into the top edge: just a thin invisible strip. Move the pointer up there and a small pill slides out; click it to open the full island. When you move away it collapses again on its own (the timing is adjustable, or set it to never close).
Double-click the pill and it hides for 30 seconds, even if your mouse passes by. Ctrl+Shift+N toggles it anytime. And if you want it fully gone, Settings β General β Show the island turns it off completely. It stops animating and uses no CPU until the shortcut or the tray brings it back.
Honest question, the approval step is nice in theory but with three subagents running in parallel I'd expect to get pinged for OK after OK. How do you keep that from turning into approval fatigue where people just mash yes? Is there a way to pre-approve low-risk stuff like reading and drafting but still gate anything that sends or spends?
@galdayanΒ yeah, that was the thing I worried about most too. If every card looks the same, people stop reading them.
There's an Auto-review mode for that. Reads, read-only commands and spinning up subagents just run, so three subagents doing research and drafts mostly won't ping you. Anything that writes a file, sends something, clicks on screen or spends still stops and asks.
You can also tell Tame in chat "don't ask me for X again", or hit "Always allow" on a card. That saves a narrow rule (one command, one folder, one host), not a blanket yes.
A few things never get pre-approved: money, security stuff, messages to your clients and logins. And if Tame read a web page or something from a connected app during that task, any send in the same task waits for you, even on Auto-review. That way a sketchy page can't slip one past you.
What would you put on the pre-approved list by default?