Your AI Coworker, Without the Landlord

The idea

Imagine you had an assistant who kept working after you left the office. Not just a chatbot you have to babysit — an actual coworker with their own computer, their own files, their own browser. You could send them a message on Slack, close your laptop, and come back to find the work done.

That's roughly what OpenAI announced in October with something called Dots. But Dots is closed, US-only, and you can only use OpenAI's own models. A small team called CopilotKit shipped an open-source version — OpenDots — the same week.

Why this one's worth watching

The big difference: you own it. You run it on your own server. You choose which AI model powers it. And every time your Dot tries to do something significant — touch a file, run a process — it stops and asks for your approval first. That's not the default with most AI tools. It's the default here.

It went trending on GitHub within hours of launch, which usually means other builders noticed something genuinely well-made.

What it means for your business

This is early. You wouldn't hand it your payroll folder tomorrow. But the pattern it represents — AI agents that persist, that work in the background, that live inside your tools like Slack — is where a lot of this is heading. Getting familiar now costs nothing.

Words worth knowing

Open-source — The blueprints are public. Anyone can inspect, copy, or improve the software. Opposite of a black box.

Self-hosted — You run the software on your own server, not on someone else's cloud. You control the data.

Model-agnostic — Works with different AI brains, not locked to one company's.

Deny-by-default — The system says no to everything unless you've explicitly said yes. A safety posture, not an afterthought.

If you have a task that repeats, takes time, and lives somewhere like Slack or email — that's the kind of thing this generation of tools is being built for. Worth a look: https://www.copilotkit.ai/blog/introducing-opendots