Cursor 3.7: Cloud Environment Setup and Cloud Subagents

Automated cloud environment setup

Cursor 3.7 brings cloud environment configuration into the Agents Window, letting developers spin up a fully functional cloud dev environment in under 10 minutes. The process is observable: a shared terminal shows each step as the agent installs dependencies, configures toolchains, and validates credentials in real time. That removes the manual friction of bootstrapping a cloud environment from scratch.

Once setup completes, Cursor captures an environment snapshot that records the installed state of the machine, including packages, system dependencies, and tooling. Future cloud agents started by any teammate can boot directly from the snapshot rather than repeating setup, cutting startup latency sharply. Teams version-control their environment definition through a .cursor/environment.json file at the repo root, making setup reproducible across everyone.

Cloud subagents with /in-cloud

The /in-cloud command introduces a new class of execution: isolated cloud subagents running in dedicated virtual machines and Git branches. They operate independently from the local session, so developers can delegate long-running work without blocking their primary workspace.

Practical uses include:

• Fixing CI failures without interrupting active local coding
• Running deep codebase exploration that would time out locally
• Preparing pull requests for merge with the /babysit command, which watches a PR until it is ready to land

Cloud subagents appear in the Agents Window sidebar alongside local agents in one unified view, and Cursor can run multiple cloud agents in parallel to scale across independent tasks.

Seamless local-to-cloud handoff

Cursor 3.7 strengthens the handoff between local and cloud execution in both directions. A session that begins locally can move to the cloud and keep running after you close your laptop. A cloud session can be pulled back into the local editor for hands-on review, testing, and iteration, then sent back to the cloud. The local and cloud boundary becomes transparent in practice: you start work wherever is convenient and shift execution context without losing state.