Cursor background agents and the CLI

What it does

• Cursor CLI. Run the Cursor agent from the terminal to write, review and modify code, interactively or for scripts and CI. How: curl https://cursor.com/install -fsS | bash, then run agent or agent "prompt" source
• Non-interactive mode. Print mode for scripts, CI pipelines and automation, with model choice and output formats. How: agent -p "find and fix performance issues" --model "gpt-5" --output-format text|json source
• File modification in scripts. In print mode the agent only proposes changes unless --force (or --yolo) is set, which applies edits without confirmation. How: agent -p --force "task"; set CURSORAPIKEY for scripts source
• Cloud Agent handoff. Push a CLI conversation to a Cloud Agent to keep running while you are away, then pick it up on web or mobile at cursor.com/agents. How: Prepend & to a message in the CLI source
• Cloud Agents. Agents in isolated cloud VMs with full development environments; run many in parallel, build and test changes, use computers to control desktop and browser, use MCP servers and work across multiple repos. How: An admin connects GitHub, GitLab, Bitbucket Cloud or Azure DevOps; start from the editor, web, iOS or integrations source
• Bugbot. Reviews pull request diffs for bugs, security issues and code quality, leaving comments with fix suggestions and Fix in Cursor / Fix in Web links. How: Configure in Automations; runs on each PR update or manually by commenting cursor review or bugbot run source
• Project Rules. Version-controlled .mdc rule files in .cursor/rules, scoped by path patterns, invoked manually or included by relevance; rule contents go at the start of model context. How: Add .mdc files under .cursor/rules source
• AGENTS.md. Agent instructions in plain markdown as a simple alternative to .cursor/rules; User Rules are global and Team Rules are managed from the dashboard on Team and Enterprise plans. How: Add AGENTS.md to the repository source
• Environments API. Beta API to programmatically create and manage the cloud environments cloud agents run in; environments support up to 300 repositories. How: Cursor Environments API (beta) source

Modes

• Agent. CLI default mode with full access to all tools for complex coding tasks. source
• Plan. Design the approach with clarifying questions before coding; Shift+Tab, /plan, --plan or --mode=plan. source
• Ask. Read-only exploration without changes; /ask or --mode=ask. source
• Print mode (-p, --print). Non-interactive scripting; add --force to apply file changes. source

Pricing and limits

• Free: $0 /month. Limited agent and Chat requests source
• Plus: $8/mo. Base usage of Cursor models (Grok 4.7, 4.6, 4.5, Composer 2.5) at fixed settings; no Auto, Cloud Agents, Automations, Bugbot, third-party models or own API keys source
• Premium: $30/mo. 4x Plus usage; adds Cloud Agents and Automations, Origin code hosting, Bugbot, third-party models at API list price from prepaid credit source
• Super: $100/mo. 15x Plus usage, priority model access at peak times source
• Ultra: $200/mo. 50x Plus usage, earliest access to new products and models source
• Teams: $30, $100, or $200 per seat/mo. Usage matches seat tier; third-party models at API list price plus the Cursor Token Rate on eligible models source
• Grok 4.7 additional usage: $2 input / $0.5 cache read / $6 output per 1M tokens. Applies beyond the weekly pool; Fast is $4 / $1 / $12 source
• Composer 2.5 additional usage: $0.5 input / $0.2 cache read / $2.5 output per 1M tokens. Fast is $3 / $0.5 / $15 source

Recent changes

• 2026-10-16: Cloud agents can build shareable interactive canvases; Environments API in beta; environments support up to 300 repositories and large ones load 5x faster. source
• 2026-10-13: Origin App Marketplace with 20+ apps and public submissions. source
• 2026-10-06: Remote control lets you see and reply to local agents from the Cursor iOS app; on by default except for Enterprise. source

Traps

• In headless print mode without --force, the agent only proposes changes and does not modify files. source
• Bugbot skips PRs larger than its default review-size limit (changed lines and diff characters) and asks you to split them; files marked linguist-generated are excluded. source
• Bugbot caps reviews on PRs opened by bot accounts per team, and every review counts including repeat reviews; it ignores cursor review / bugbot run comments posted by bots or GitHub Apps. source
• The Plus plan excludes Cloud Agents, Automations, Bugbot, third-party models and your own API keys; these start at Premium. source
• Third-party model usage is not included in plan prices and is paid separately from prepaid credit at API list price. source
• Older Pro, Pro+, Ultra and Teams Standard/Premium seats are legacy plans with two monthly pools and on-demand usage, while current plans use one weekly-refreshing pool, so usage advice differs by plan. source

AuthorAndrey Reshetnikov, AI product and AI engineering