Summary

Cursor shipped 'Self-hosted machines' (9/2 changelog): tool execution for cloud agents stays entirely on the customer's own network — codebases, build artifacts, and secrets remain on internal infrastructure. Scheduling is a dynamic pool: personal 'My Machines', scalable team pools, and idle-worker hibernation. Cloud agents can run on existing sandboxes — AWS Lambda, Coder, Cloudflare, Daytona, Modal, Namespace, Vercel, E2B — and computer use works on self-hosted Linux and Mac machines.

Why it matters
Always-on cloud agents have been the part of the coding-agent stack that data-residency-constrained enterprises could not adopt, because execution meant sending code to vendor infra. This turns the execution plane into bring-your-own infrastructure, reusing sandbox providers teams already run. It pairs with OpenAI's ZDR/Private Safety Processing commitments and Anthropic's Enterprise Frontier Safeguards as the enterprise control surface assembles (candidate trend).
Technical details
Execution tool execution on the customer's own network; codebases, build artifacts, secrets stay internal
Scheduling dynamic pool: personal 'My Machines', scalable team pools, idle-worker hibernation
Sandbox Providers AWS Lambda, Coder, Cloudflare, Daytona, Modal, Namespace, Vercel, E2B
Computer Use supported on self-hosted Linux and Mac machines
Tags
cursorself-hostedcloud-agentsdata-residencysandboxcomputer-use