Today we are releasing Cloudflare OS, a chatbot with connectors, just like every other tech company is doing.
Except actually, it's different. This is a remake of Sandstorm[.]io, my startup from 10 years ago, except this time built on Cloudflare Workers (the platform I've spent the last 9 years building) and deeply leveraging AI. This is more or less the culmination of my secret 10-year master plan.
This is a full-on personal app vibe coding platform, in which the sandbox is so secure that you can pretty much go wild -- the AI cannot introduce a significant security bug. We believe a company's security team can feel comfortable giving non-technical users permission to vibe code and then sleep soundly at night.
How is that possible? It's the Sandstorm security model, revisited. A "Gadget" is the same thing as a Sandstorm "Grain": a fine-grained app instance. For example, if you have a document editor app, each document runs as a separate instance of the app, in a separate sandbox (one "Gadget").
This means two things, both of which I think are Big Deals:
1. The platform can manage all access control, by controlling who can access the Gadget at all. There is no way the Gadget can accidentally leak itself to an attacker -- even an attacker who has access to other Gadgets based on the same app.
2. Since everyone is running their own copy of the code, everyone can freely *modify* their copy of the code.
Think about #2 a bit more.
What if, when you wanted a new feature in the software you are using, you could just prompt your agent to add it?
This doesn't work in the cloud Software-as-a-Service model, because you are not running your own copy of the app.
Sandstorm tried to change that 10 years ago, but the world wasn't ready, because not enough people had the skills or patience to actually modify their software. AI has changed that. Now you just ask the agent -- the same agent that you are using to help you interact with the Gadget can also modify the code of the Gadget.
And it is so fun.
https://github.com/cloudflare/cloudflare-os…
Of course, personal apps are more useful if they can connect to external services. Cloudflare OS introduces a "connector" system we call Gatekeepers. This is sort of like MCP (and MCP is supported as a kind of Gatekeeper), but with a lot more:
* Instead of exposing tools, a Gatekeeper exposes a Cap'n Web RPC API. That makes it appropriate for use by both agents (via code mode) and Gadgets.
* Gatekeepers integrate with the Cloudflare OS UI to provide inline audit logging and human-in-the-loop approvals for all side-effecting actions.
* When an action requires approval, the agent does not need to stop and wait for it. A Gatekeeper will *simulate* the outcome, allowing the agent to keep running and queue up more work. You can then approve everything in a batch at the end. Hopefully, this means you no longer feel the need to turn on auto-approve! (But you still can if you want.)
We have already built Gatekeepers for a huge number of services, from GitHub to Home Assistant. We've found, with the right skills, AI can basically crank these things out for any given API, solving the chicken-and-egg ecosystem problem. We are working on a "software factory" of sorts where you can input the API docs for your favorite service and get a well-written Gatekeeper. That's not quite ready yet -- so for now use the ones we provide or use the write-gatekeeper skill to write your own. You can add your own Gatekeepers to your personal deployment.
I am so proud to say, we are making this whole thing open source, Apache 2.0 license, no catch.
Truth be told, we weren't quite ready to launch this yet. That's why there's no online version you can just go to yet. But rather than pre-announce vaporware, we decided to go ahead and open up the repo now. We even built a little tool to help you get it all deployed on your own Cloudflare account.
No lock-in, though: You can also self-host it on workerd, the open source Workers Runtime. This can run on your own machine on a 100% open source stack.
Why make it open source? Well, frankly, if we didn't, someone else would just slop-clone it. But also, we think every company (and enthusiast) will want to run their own instance of this, connected to their own internal services, and with their own customizations. Cloudflare OS is the instance Cloudflare runs, but your instance is *your* AI OS. It's so much better this way.
https://github.com/cloudflare/cloudflare-os…
Here's the announcement blog post. Warning: it's enterprise-focused. Let's be frank, that is the biggest market opportunity. But I personally very much care about the indie use case as well. I'm running an instance of this in my own basement.
Speaking of which, the Gadget platform is so simplified compared to "normal" cloud environments that local LLMs can actually handle coding for it reasonably well. Not perfect, but you just feed back the errors to them and they eventually figure it out. And they keep getting better. Of course, you can also run it locally while connecting to Claude or GPT for inference, if you prefer.
https://blog.cloudflare.com/cloudflare-os/
Also our landing page is pretty sweet, but as of this tweet doesn't have OG card info yet so doesn't render a nice image, oops. Workin on that.
https://os.cloudflare.app/…
A Cloud for Small Software
@Pete Koomen
Agents make it easy to build personal tools for yourself or your team. But deploying, securing, and sharing that software is still far more complicated than creating it.
A cloud built for small software could remove that complexity and make bespoke tools as easy to share with a colleague as a Google Doc.
I should mention, Cloudflare OS is the combination of several ideas from different parts of the company. I dwell on the Sandstorm lineage here because it's obviously what I'm excited about.
However, there's a whole other side that's about making AI work well for your company, making sure it has the right context to do things the way your company does things. This was developed by @sam rhea, @🤸🏽, and others as Cloudflare OS v1, in parallel with my work.
It was just about a month and a half ago that we merged the projects -- taking the ideas from Cloudflare OS v1 and putting them on top of the core of "Gadgets", the Sandstorm successor, to make one product, Cloudflare OS v2.
We have compared this to Windows 95 vs. NT. Windows 95 was a revolution in UX. It was later retrofitted on top of the NT kernel to create Windows 2000.
Cloudflare OS v1 is used widely within Cloudflare already, with v2 only beginning to replace it. v2 is what we're releasing, and it still has a lot of rough edges. We are roughly at NT 4 now, working towards 2000 and then XP.
BTW, this past fun tweet was on an early prototype of this project. This was sort of an emergent feature I didn't even intend:
A Gadget is a combination of some client code and some server code connected over Cap'n Web RPC, so that the client can just call methods on the server without any fuss.
Meanwhile, the agent performs actions by writing and executing snippets of code. ("Code Mode") External resources are exposed to the agent as Cap'n Web RPC interfaces for this purpose.
I enabled the agent to directly call the RPC interface of the Gadget it was building.
And suddenly, I had an agent that could collaborate with me on *any Gadget*. Since it can just look at the code, see the RPC interface, and start calling it. And all Gadgets are inherently real-time collaborative (Durable Objects and Cap'n Web make it so easy), so whatever the agent did would show up on my screen in real time.
https://x.com/KentonVarda/status/2029678769372299508…
Opus 4.6 is smart enough to play tic tac toe on this whiteboard with me entirely by making API calls to the app's client API, yet dumb enough to lose at tic tac toe.
Today we're open sourcing Cloudflare OS:
- An agent workspace grounded in skills your team curates.
- A platform for personal, modifiable apps that people can build and share.
- A new security framework for safe access to internal data.
You can use it to: