Cursor CLI Non-Interactive Mode (Scripts & CI)

Interactive mode: the AI asks you questions, you press Y/N, and confirm step by step — like pair programming with someone. Non-interactive mode: give it one instruction, it runs to completion, no one needs to sit there pressing keys. Perfect for scripts or CI pipelines — let the AI work overtime while you do something else.

Basic Usage

Add --no-interactive (or the shorthand -n):

cursor agent "Add a .editorconfig to the project" --no-interactive

The AI executes directly without pausing for input
Best for: predictable, repeatable tasks (adding comments, fixing lint, reformatting to standards, etc.)

Using in a Script

#!/bin/bash
cd /path/to/project
cursor agent "Run the linter and fix issues per CONTRIBUTING.md" --no-interactive

Works for automated code cleanup, simple refactors, adding comments
Remember: the CI environment needs Cursor account authorization (run cursor auth login first, or use a token) — otherwise the AI can't do anything

Specifying a Model (Optional)

Want to use a specific model for the run:

cursor agent "Rewrite this comment" --model claude-3-5-sonnet --no-interactive

Check the official docs for available model names — don't make up model IDs.

Run in the Cloud (-c)

Heavy task and don't want to tie up your terminal? Use -c to hand it off to the cloud:

cursor agent "Analyze the entire repo and generate an architecture document" -c

-c hands the conversation off to a cloud Agent to continue
You can close your terminal — you can reconnect to the conversation later. Think of it as "background execution," but it's the AI working in the background for you.

Summary

Non-interactive: add --no-interactive or -n, no one needs to press keys
Scripts/CI: same command — just make sure the environment can log in to Cursor
Specify a model: --model model-name
Long-running tasks: use -c to hand off to the cloud, works even if you shut down

Next: 05-tips — Tips and FAQ to avoid common pitfalls