Open-source OpenAI Dots alternative opendots hits 2.8K stars in two days
opendots is an MIT-licensed multi-assistant framework that runs locally, giving each dot its own browser, files and permissions, requiring human approval before saving, and supporting Slack and voice.
A REPO WENT FROM 1.3K TO 2.8K STARS IN TWO DAYS. IT'S A FREE, OPEN-SOURCE VERSION OF OPENAI'S DOTS it's called opendots, mit licensed, runs on your own machine not one chatbot - a team with jobs: → gives each dot a name, a role, and its own list of allowed tools - one researches, one writes, separate conversations, you control exactly what each one can touch → hands each dot its own computer - its own browser, files and terminal, profile and files survive restarts, a live panel to watch it work, you can take the browser back any second → pauses before it saves anything - ask for a draft and the chat stops on a card with 2 buttons, approve or decline, nothing becomes a page until you click approve → turns approved work into pages, not a chat log - spaces: a searchable library, visual editor, slash commands, nested subpages, autosave → answers by voice while it keeps working - live captions, mute, a call timer, a second agent runs the long task in the background while you talk → shows up where you already work - mention a dot in a slack thread, it keeps working there, an allowlist controls which workspace and users can reach it → runs on a schedule and gets better without you - background tasks with pause and retry, automatic learning turns patterns from your conversations into reusable skills setup is 4 commands: clone it, npm ci, copy the env file, npm run dev most multi-agent clones give every agent the same shared browser and call it a team. this one gives each dot its own profile, its own files, its own permissions - that's the line between a demo and something you'd actually trust with a task while you're not watching i track every repo in this dots wave so you don't have to. follow @0xkkai if you want the next one before it hits 10k so here's the real question: if anyone can clone a team of always-on agents in four commands, what exactly is openai charging for?
This post quotes another post: x.com/teka1900/status/2105602481506156562