# Cursor Memories Memories let [[Cursor.com|Cursor]] keep facts from one conversation and reuse them later, scoped to a project. The agent remembers decisions, conventions, and context you would otherwise have to repeat every session. Memories shipped with Cursor 1.0 in June 2025. ## How it works - Memories are stored per project, not globally, so one repo's context does not leak into another - The agent can form a memory on its own during a conversation, and you can review and edit what it kept - Stored memories feed back into context on later agent runs in that project ## Notes - Memories cut repeated prompting, but they are only as good as what the agent chose to keep. Worth reviewing now and then. - This is the opposite of [[Cursor Rules]]: rules are written by you and always apply, while memories are accumulated by the agent over time. ## References - Changelog: https://cursor.com/changelog ## Related - [[Cursor.com]] - [[Cursor Rules]] - [[Cursor Agent Mode]] - [[Claude Code Memory]] - [[AI Agents]]