Cursor’s command line interface uses the agent command. It is useful when the task belongs in the terminal or when you want a scriptable agent surface.

Interactive Sessions

Run agent inside a repository:

agent

The terminal interface supports Agent, Plan, and Ask modes. It can resume sessions, continue the latest session, and hand work to the cloud when you prepend & to an agent request:

& refactor the auth module

Useful session commands include:

agent ls
agent resume
agent --continue

Print Mode

Print mode runs a prompt and exits:

agent -p "Summarize the staged diff and list test gaps."

By default, print mode is safest for read-only work. If you allow file modification, do it explicitly with the documented flags and only inside a clean working tree.

Automation Pattern

Use print mode for narrow checks:

agent -p "Review this diff for missing tests. Output only findings with file paths."

Do not build hidden release systems out of agent prompts. If a step must always run, put it in a script or continuous integration. Use the agent to reason about the result.

Local Permissions Still Matter

The command line interface can run shell commands and edit files. The same approval rules apply as in the editor: know what the agent is allowed to do, review the diff, and run the verification command yourself before shipping.