United States – Cloudflare has announced support for running Cursor Cloud Agents on Cloudflare Sandboxes, allowing developers and platform teams to execute AI coding agents in isolated environments managed within their Cloudflare accounts.
The integration extends Cursor’s Self-Hosted Machines workflow, which allows users to run Cursor agents on customer-selected infrastructure while continuing to manage those agents through the Cursor application, website and mobile app.
Under the arrangement, Cursor continues to handle the agent loop, including inference, planning and orchestration, while Cloudflare Sandboxes provide the environment in which agent tasks are executed. Tool calls such as terminal, filesystem and browser operations run within the customer-controlled sandbox.
Cloudflare said the approach is intended to give organisations greater control over where agent workloads run and how they access code, systems and secrets.
The company has also highlighted integrations with other AI agent platforms, including Devin Outposts and Claude Managed Agents, as part of its wider development of Cloudflare Sandboxes as an execution environment for agent-based applications.
Cursor Cloud Agents use a self-hosted worker model in which the Cursor CLI establishes a long-lived outbound HTTPS connection to Cursor’s backend. This allows agent tool calls to be sent to the worker without requiring an inbound connection into the customer’s network.
“Developers want powerful AI tools that fit naturally into their workflows, and enterprises need those tools to run in environments they control,” said Dane Knecht, Chief Technology Officer at Cloudflare.
“Cloudflare Sandboxes gives teams the freedom to use the AI tools they prefer while giving organisations a secure, isolated place to run agent work. Bringing Cursor Cloud Agents to Cloudflare Sandboxes is another step toward making Cloudflare the execution layer for the next generation of agentic applications,” he continued.
The integration supports both individual developers and enterprise teams using Cursor Self-Hosted Machines. Individual users can connect a worker through My Machines, while organisations can use self-hosted worker pools as routing targets for agent sessions.
Worker pools can be configured for different execution environments, with pool orchestration allowing teams to monitor demand, add worker capacity when required and release capacity after sessions end.
“SpaceXAI’s goal is to make agents useful wherever developers work. Self-Hosted Machines let teams keep Cursor workflows across desktop, web, and mobile while routing work to infrastructure they operate. Cloudflare Sandboxes give those teams an enterprise-grade option for running agent workloads in a controlled environment,” said Toni Adams, Senior Director of Partnerships at SpaceXAI.
During self-hosted operation, repositories, build caches and secrets remain on the customer’s machines. File chunks accessed by the model during inference, as well as Cloud Agent artefacts including screenshots, videos and log references, are uploaded so they can be displayed in pull requests and dashboards.
The setup also supports the Cloud Agents API, which organisations can use to integrate self-hosted machine status and worker pool routing with their own systems.
The announcement comes as enterprises increasingly explore ways to deploy AI coding agents while retaining control over the infrastructure and data used by those systems. Cloudflare said its Sandboxes provide an additional execution option for teams already using Cursor Self-Hosted Machines across different environments.
