AI Agents Can Now Spin Up Cloud Apps Without Human Help
Cloudflare and Stripe have launched a new protocol that lets AI agents create Cloudflare accounts, start paid subscriptions, register domains, and deploy code—all without a human ever touching a dashboard. Starting today, an agent working on behalf of a person can go from zero to a live web app in a single automated sequence.
Human users must first accept Cloudflare's terms of service, but after that, their involvement becomes optional. The AI handles everything: account creation, payment setup, domain registration, and API token delivery. Cloudflare product managers Sid Chatterjee and Brendan Irvine-Broque describe it as “one-shot deployment.”
While developers and startups may celebrate the speed, security experts warn this lowers the barrier for threat actors. David Shipley, CEO of Beauceron Security, told The Cyber Wire: “Cybercriminals are constantly building and discarding infrastructure to evade takedowns. Making it even faster to build new infrastructure and deploy it quickly is a huge win for them.”
OAuth Keys Handed to Agents: A New Authentication Flow
The protocol, co-designed by Cloudflare and Stripe, builds on Cloudflare's Code Mode MCP server and Agent Skills. Any platform with signed-in users can integrate it with “zero friction” for the user, according to the companies' joint blog post. Agents receive an initial $100 monthly budget per provider via Stripe Projects, still in beta.
To use the system, a human installs the Stripe CLI with the Stripe Projects plugin, logs in, and starts a new project. The agent then prompts a build command. If the user's Stripe login email matches an existing Cloudflare account, an OAuth flow kicks off automatically. Otherwise, Cloudflare creates a new account on the spot.
From there, the agent builds and deploys a site to a fresh Cloudflare account, uses the Stripe Projects CLI to register a domain, and configures the app to run on that domain. The agent asks for input only when necessary—for example, if no payment method is linked. Cloudflare says it goes from “literal zero to full deployment”.
Background: From Manual Setup to Full Autonomy
Previously, deploying a new web application required humans to manually sign up for cloud services, enter credit card details, copy API tokens, and configure DNS settings. This process could take hours or even days. Cloudflare and Stripe's new protocol eliminates nearly all manual steps, giving agents the ability to provision infrastructure on demand.
The move follows a broader industry trend of “agentic workflows,” where AI models execute multi-step tasks. However, until now, most cloud platforms retained a human-in-the-loop for billing and account creation. Cloudflare is the first major cloud provider to grant full autonomy to agents for all provisioning steps.
To spur adoption, Cloudflare is offering $100,000 in credits to startups that use the capability through Stripe Atlas, a service that helps companies incorporate in Delaware, set up banking, and raise funds.
What This Means: Speed vs. Security Tradeoff
For legitimate developers, the ability to spin up infrastructure in seconds could accelerate prototyping and reduce friction. Startups using Stripe Atlas can now deploy apps without waiting for human approvals. But the same speed benefits cybercriminals, who constantly recycle infrastructure to avoid detection and takedowns.
Security researcher David Shipley added: “This is exactly the kind of automation fraudsters have been dreaming of. They can now script a botnet's entire hosting lifecycle.” Organizations must reconsider their zero-trust postures: if an agent can create accounts, who—or what—controls the API keys?
The protocol itself includes OAuth approval at the initial token exchange, but once granted, the agent has full access until the token expires. Cloudflare says agents will prompt for approval “when necessary,” such as when no payment method is linked. However, critics argue that an agent could, in theory, chain multiple services together without human intervention, creating a blind spot for governance.
Enterprise security teams are now urged to review any integrations that grant agents autonomous billing and domain registration capabilities. As one industry analyst put it, “We're handing the keys to the cloud to machines that don't have ethics or legal liability.”