Cursor shipped a CLI update that adds agent modes and cloud handoff directly to the terminal. Developers who split time between the editor and the command line no longer need to context-switch back to the GUI for Cursor’s most capable AI features.
What happened
cursor-cli now supports three distinct agent modes — the same ones previously locked to the desktop editor. The CLI can launch foreground agents that work interactively in your terminal session, background agents that run tasks autonomously while you keep working, and cloud-based agents that hand off long-running jobs to Cursor’s servers entirely.
Cloud handoff is the most consequential addition for workflow flexibility. You can kick off an agent task from your laptop, close the terminal, and pick up the results later from any device. The agent runs on Cursor’s infrastructure, not your local machine. For CI-adjacent workflows or large refactors that take minutes to complete, this removes the “keep your laptop open” constraint.
The update also includes improved shell integration and better streaming output for agent responses.
TIP
Cloud handoff agents persist across sessions. Start a refactor from your terminal at work, check results from your home machine. No editor window required.
Why it matters
A significant chunk of professional development still happens in terminals — SSH sessions, remote servers, CI pipelines, tmux setups. Until now, using Cursor’s agent capabilities meant opening the desktop app. That friction pushed terminal-heavy developers toward alternatives like Claude Code or Aider that were CLI-native from the start.
This update addresses that directly. Background agents in the CLI mean you can run cursor-cli agent in one tmux pane while coding in Neovim in another. Cloud handoff goes further: it decouples the agent’s compute from your local session entirely, which matters for developers working on resource-constrained machines or flaky connections.
WARNING
Cloud agents execute on Cursor’s servers, which means your code context gets sent to their infrastructure. Review your team’s security and compliance requirements before using cloud handoff on proprietary codebases.
Feature comparison: CLI vs. desktop editor
| Capability | Desktop Editor | CLI (after update) |
|---|---|---|
| Foreground agents | ✅ | ✅ |
| Background agents | ✅ | ✅ |
| Cloud handoff | ✅ | ✅ |
| Streaming output | ✅ | ✅ (improved) |
| Shell integration | N/A | ✅ (improved) |
| Works over SSH | ❌ | ✅ |
| Runs headless (no GUI) | ❌ | ✅ |
Who should care
Terminal-first developers who already use Cursor’s editor but want agent capabilities in their CLI workflow. DevOps engineers running agents against infrastructure code in SSH sessions. Anyone who has tried Claude Code or Aider specifically because Cursor lacked a capable CLI.
Teams using Cursor Business plans should also pay attention: cloud handoff could fit into automated code review or migration pipelines where you want an agent to process a task without tying up a developer’s machine.
NOTE
The CLI agent modes use the same underlying model access as the desktop editor. Your existing Cursor plan and model availability apply — no separate subscription needed.
Practical takeaway
Update your Cursor CLI to the latest version and try cursor-cli agent --mode background on a contained task — a test file generation or a single-module refactor. Test cloud handoff on a non-sensitive repo first to understand latency and output quality before integrating it into real workflows. If you already have agent prompts that work well in the GUI, they are worth testing in the CLI to see how well they carry over — the feature set maps closely to the desktop editor, though exact parity is not guaranteed for every edge case.