Cursor Background Agents: Complete Guide to Async AI Coding

Cursor Background Agents are autonomous AI-powered assistants that execute coding tasks independently in isolated Ubuntu cloud VMs while you continue working locally. ## Getting Started Hit Ctrl+E to trigger background agent mode in the UI. Write your prompt describing the task. The agent clones your repo into a fresh VM, works on a separate branch, and pushes a PR when done. ## Key Features - **Asynchronous Execution**: Work on other tasks while the agent codes in the background - **Isolated Environment**: Each agent runs in a fresh Ubuntu cloud VM - **Computer Use**: Agents can interact with desktop, browser, and UI elements - **Automatic PR Creation**: Changes are pushed as pull requests for review ## Best Use Cases 1. Generating unit tests for entire directories 2. Bulk refactoring (e.g., class components to functional) 3. Migrating codebases between frameworks 4. Running long-running analysis tasks 5. Documentation generation across large codebases ## Tips for Success - Always commit to Git before running background agents - Write clear, specific prompts with explicit requirements - Review generated PRs carefully before merging - Use for well-defined tasks you can hand off entirely ## Background Agents vs Agent Mode Agent mode is interactive and local, requiring confirmation for each action. Background agents are asynchronous and cloud-based, executing without confirmation and pushing changes as PRs. Use agent mode for steering the AI interactively; use background agents for delegating complete tasks.