Cloudflare spent several months building an internal AI agent workspace called Cloudflare OS, tested it with thousands of employees, and on August 5 pushed the whole thing to GitHub under an open source license.
What It Does
Cloudflare OS lets employees describe workflows in plain language. An AI agent then codes those descriptions into working applications. According to Cloudflare, daily use cases include creating documents and slides, automating repeatable tasks, and building small data visualization apps. No engineering background required.
The Security Model
The obvious concern with non-technical staff generating live code is the blast radius when something goes wrong. Cloudflare’s answer is fine-grained sandboxing: each app runs as a separate instance, each document gets its own sandbox, and each user runs their own copy of the code.
“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.” — Kenton Varda, principal engineer at Cloudflare
The platform manages permissions per instance, so a rogue document editor app can’t touch another user’s data or another app’s runtime.
Model Flexibility and Cost Control
Cloudflare OS works with any model available through Cloudflare’s AI Gateway, and organizations can assign different models to different tasks. Sam Rhea, Cloudflare’s CIO, framed the cost angle directly: “Not every user needs access to the max thinking mode of the latest frontier lab model. And we do not need team members spending $20 to summarize their email inbox every hour.”
The open source release is available now on GitHub. If you run a team where non-engineers are already attempting to build internal tooling, this is worth a look before someone’s prompt-generated app lands in production with no sandbox around it.
