# Cursor Agent Mode Agent Mode is the mode in [[Cursor.com|Cursor]] where the AI takes the lead. You describe a goal, and the agent plans, edits across files, runs commands, and checks its own work until the goal is met, instead of waiting for you to drive each step. It arrived in v0.45 and has since become the default way most people use Cursor. ## What the agent does - Edits multiple files in one pass, and creates new ones - Runs terminal commands and reads their output - Searches the codebase for relevant context on its own - Loops: act, observe the result, correct, repeat ## Notes - Agent Mode is the foundation the rest of the agent features build on. [[Cursor Subagents]], [[Cursor Multi-Agents]], [[Cursor Cloud Agents]], and [[Cursor Automations]] are all variations on running this loop. - [[Cursor Rules]], [[Cursor Skills]], and [[Cursor Hooks]] exist to constrain and extend the loop. ## References - Documentation: https://cursor.com/docs - Changelog: https://cursor.com/changelog ## Related - [[Cursor.com]] - [[Cursor Subagents]] - [[Cursor Multi-Agents]] - [[Cursor Cloud Agents]] - [[Agentic loops]] - [[AI Agents]]