OpenWork: your AI's skills, tools, and logins, set up once and used everywhere

2 months ago

You set up one AI app. Then the next one makes you start from scratch. OpenWork lets you do that setup once, then carry the whole thing into any AI app you use. The skills you taught it, your tools, your logged-in accounts, all ready the moment you connect.

Ask

Ask about this presentation

Answers are generated from this presentation.

Chapters

  1. 0:00OpenWork: your AI's skills, tools, and logins, set up once and used everywhere
  2. 0:14A workspace where you describe a task and get finished work back
  3. 0:33You rebuild the same AI setup in every app
  4. 0:49Everything you'd give your AI becomes one kind of thing
  5. 1:17Two buttons, forever
  6. 1:50One setup, reachable from every app, machine, and teammate
  7. 2:15Free and open source, and yours to run anywhere
  8. 2:36Set up your AI once, and it follows you everywhere
Show transcript

OpenWork: your AI's skills, tools, and logins, set up once and used everywhere

GitHub Trending · Ep 24 · Jul 31, 2026

OpenWork

Your AI's skills, tools, and logins, set up once and used everywhere.
TypeScript MIT core Mac · Windows · Linux

You set up one AI app. Then the next one makes you start from scratch. OpenWork lets you do that setup once, then carry the whole thing into any AI app you use. The skills you taught it, your tools, your logged-in accounts, all ready the moment you connect.

A workspace where you describe a task and get finished work back

A workspace where you describe a task and get finished work back.

And the app is optional. Bolt OpenWork onto the AI you already use.
Hand it a task in plain words, pick a model, and OpenWork does the work.
In this run
it built a small data file, ran some code over that file, and opened the finished result for you.
Your whole setup comes along
into whatever AI app you already use.
src: OpenWork README

Here's OpenWork itself. You hand it a task in plain words, pick a model, and it goes and does the work. In this one, OpenWork built a small data file, ran some code over that file, and opened the finished result for you. And here's the key. The app is optional. You can bolt OpenWork onto the AI app you already use, and your whole setup comes along.

You rebuild the same AI setup in every app

You rebuild the same AI setup in every app.

Every app is its own island
Connect your accounts in one, add your skills in another, and none of it carries over.
More power means a longer menu
The longer that tool list gets, the slower your AI runs, and the more often it grabs the wrong one.

Right now, every AI app is its own island. You connect your Google account in one, add your skills in another, and none of it carries over. And the usual way to give an AI more power is to keep bolting on more tools. But the longer that tool list gets, the slower the AI runs, and the more often it grabs the wrong one.

Everything you'd give your AI becomes one kind of thing

Everything you'd give your AI becomes one kind of thing.

OpenWork calls each one a capability.
A skill you taught it
A connection to another tool
Your logged-in Google account
Your logged-in Microsoft account
→
One catalog of capabilities
Every one is now the same kind of thing, kept in one place that lives with you, not buried in a single app's settings.
Four unlike things become one kind of thing, so one small interface can reach every one of them.
src: OpenWork README + capability rail

So here's OpenWork's real move. OpenWork takes everything you'd hand an AI and files it all under a single label. A skill you taught it. A connection to another tool. Your logged-in Google account. Your logged-in Microsoft account. Four completely different things, and OpenWork calls every one of them a capability. OpenWork keeps them in one catalog, and that catalog lives with you, not buried inside one app's settings. And this part's clever. Because they're all the same kind of thing now, one simple interface can reach every single one.

Two buttons, forever

Two buttons, forever.

Search and run, no matter how much you add behind them.
Your AI app
Claude Code, Cursor, Codex, or any add-on-capable app.
→
One small add-on, two buttons
search_capabilities find what you can do
execute_capability run it
That's the whole interface your AI ever sees.
→
Your growing catalog
skill tool connection Google Microsoft saved skill calendar docs + hundreds more
More than a hundred operations back there. Your AI still sees just two buttons.
The catalog grows. The buttons stay at two.
src: OpenWork agent rail (search_capabilities + execute_capability)

Here's the payoff of making everything one kind of thing. Instead of handing your AI a hundred separate tools, OpenWork hands it exactly two, forever. One button to search your capabilities and find the one that fits the job. One button to run the one it found. That's the whole interface. Add a thousand more capabilities behind those two buttons, and your AI still sees only two. Under the hood, OpenWork runs well over a hundred operations. Your AI still sees just those two buttons. Everything else is searchable data sitting behind them. So what your AI can do keeps growing, while the menu it has to reason about never grows past two. The abilities pile up. The buttons stay put.

One setup, reachable from every app, machine, and teammate

One setup, reachable from every app, machine, and teammate.

Any AI app you already use
Claude Code Cursor Codex Any add-on-capable app
←
Your one capability catalog
Lives with you, not inside any single app.
→
Everywhere you work
Every machine you own Your teammates
Set it up once, and it follows you.
src: OpenWork README · Den-served catalog (Claude Code / Cursor / Codex / any add-on-capable app)

And because that catalog lives with you, not inside one app, the same setup is reachable from whatever AI you already use. Claude Code, Cursor, Codex, or any app that takes an add-on like this. Bolt it on once, sign in, and your whole setup is just there, on your laptop and your desktop. Hand it to a teammate, and they get the same abilities you set up. And a team gets one shared place to manage who can use what. Set it up once, and it follows you.

Free and open source, and yours to run anywhere

Free and open source, and yours to run anywhere.

Download the desktop app
Nothing to configure. Runs on Mac, Windows, and Linux.
Or bolt it onto your AI app
Paste one line, then sign in, and your setup comes along.
claude mcp add --transport http openwork \ https://api.openworklabs.com/mcp/agent
The main app is free for anyone, even inside a business. The enterprise extras are source-available today, and turn fully open after two years.
MIT core v0.18.12 Mac · Windows · Linux
different-ai/openwork · README

Two ways in, and both are free. Download the desktop app for Mac, Windows, or Linux. Or, if you already use an AI app, paste one line to bolt OpenWork on, then sign in. It's open source. The main app is free for anyone, even inside a business. The enterprise extras are source-available today, and turn fully open after two years. It's still early, at version zero point eighteen, and shipping fast.

Set up your AI once, and it follows you everywhere

OpenWork logo, an isometric stacked-cube mark

Set up your AI once, and it follows you everywhere, because everything is a capability behind two buttons.

github.com/different-ai/openwork
New repo tomorrow.

You don't set your AI up again in every app. You set it up once, and one small add-on carries the whole thing into any AI you use, and out to your team, because everything is a capability behind two buttons that never multiply. The link's on screen, and there's a new repo tomorrow.