It’s not really an operating system, but an open-source platform for agentic AI workspaces.
So, you want an open-source platform to develop and run AI agents? Cloudflare, the well-known content delivery network (CDN) company, continues to move beyond its foundations to deliver its internal AI tooling, Cloudflare OS.
Originally built to run the company’s own global workforce, the platform is now positioned as an AI “operating system” for enterprises: a browser‑based workspace where employees can research, generate documents, build micro‑apps and automate workflows, all wired into internal systems and governed by Zero Trust security policies.
Cloudflare co‑founder and CEO Matthew Prince said, “Cloudflare OS is how we run Cloudflare. For AI to truly transform an enterprise, it can’t live in a silo or behind a developer bottleneck. Every employee needs the ability to build, iterate, and automate safely.”
That all sounds great, but what does it actually do? According to the company, despite the name, Cloudflare OS is not an operating system in the traditional kernel‑and‑userspace sense. Instead, it’s “an open platform for agents, apps, and work,” and an “operating system” in two metaphorical ways: one for how a company safely becomes productive with AI, and another for how AI workloads themselves are managed, much as a conventional OS manages processes.
At its heart, the program is an AI agent workspace accessed through the browser. Each conversation is grounded in organization‑specific context and “skills.” The latter are curated connections to documents, APIs, and internal systems, which are backed by an isolated runtime where agents can write and execute code. The platform is tightly integrated with the company’s Zero Trust service, Cloudflare Access. Under it, every user and every agent request is authenticated and authorized before being allowed near corporate data or services. That meets my approval.
Cloudflare OS also blurs the line between chat, apps, and automation. AI session outputs can be turned into micro‑applications with their own isolated database, real‑time capabilities, and access controls. Your co-workers can then use, clone, or adapt these to their needs. Workflows can also be set to run in the background, effectively treating agents as long‑lived processes rather than one‑off completions.
In keeping with Cloudflare’s recent push for model‑agnostic AI infrastructure, the platform supports “bring your own model” via Cloudflare AI Gateway. This enables you to route workloads to your model provider of choice, instead of locking into a single vendor.
The platform is aimed squarely at enterprises that want to democratize AI inside the business without handing every initiative to a central development team. Cloudflare pitches OS as “an agent workspace for everyone,” emphasizing that employees can build and share tools “no developer required” while security and governance teams retain tight control over what data agents can see and what systems they can touch.
Cloudflare frames it as a way to capture “working knowledge” from day one. The patterns encoded in micro‑apps and workflows are meant to persist as employees move roles or leave, rather than living only in personal scripts or ad‑hoc documents.
I find this approach more than a bit worrisome myself. It’s not that I don’t trust non-tech-savvy people with AI tools; it’s that I’ve seen far too many times when they’re overwhelmed and make blunders that an experienced developer would never make. Still, with the right guidelines, it may work well for your group.
Cloudflare’s decision to open‑source Cloudflare OS aligns with its broader campaign to present itself as a connectivity cloud and an AI infrastructure layer rather than a closed SaaS stack. By shipping Cloudflare OS under Apache 2.0, the company is explicitly promising that organizations will own the workflows, micro‑apps and system integrations they build on top of it, rather than entrusting them to a proprietary black box.
That said, Cloudflare OS runs on top of its own global network. More specifically, it runs on the open‑source Cloudflare Workers runtime (workerd). You can deploy this as a managed service on Cloudflare’s connectivity cloud or self‑host it on workerd on your own servers. Still, at day’s end you’ll need to work with Cloudflare’s network to make the most of it.
You may be asking yourself, “What else is like this?” The answer is: “Nothing already out there really.” You see, Cloudflare OS competes across three overlapping buckets: enterprise AI workspaces, agentic automation platforms, and AI‑enabled connectivity/Zero Trust stacks. The closest “comparables” are products like Microsoft Copilot for Microsoft 365, Google’s Workspace/Gemini combo, agentic automation platforms such as UiPath Agentic Automation, and AI operating‑system style platforms like AIOS.
So, who is this for then? I see three possible markets. First, and the most obvious, organizations already using Cloudflare’s network, security, or Zero Trust services and looking for a native way to layer AI on top. Another is companies with complex internal systems, such as from legacy databases to SaaS applications, that want a single agentic fabric to connect them, with fine‑grained permissions and auditability. Finally, IT and security teams that are under pressure to enable AI experimentation across the business without compromising control or compliance may find it worthwhile after they’re content with its guardrails.
Cloudflare OS is launching into a larger narrative the company is building around an “agent era” for cloud infrastructure. Just days before the OS announcement, Cloudflare previewed `@cloudflare/computer`, an open‑source agent runtime library that orchestrates where and how agent code executes across isolates, container sandboxes, or the browser. The open sourcing of Cloudflare OS adds the human‑facing workspace, governance layer, and application framework on top of that technical foundation.
For Cloudflare, the bet is that enterprises will want a consistent way to wire agents into their existing networks and security posture, and that making the core platform open source will help build trust in how that wiring works. I think they have a real shot at finding customers, especially since anything else remotely like Cloudflare OS will require you to either lock yourself into a Microsoft or Google AI platform or have your in-house IT staff invest blood, sweat, and tears into customizing a platform for your needs.
