Claude Code has introduced the "Auto Memory" feature. Anthropic has launched the Auto Memory feature for its command-line coding tool, Claude Code. Once enabled, Claude will automatically record the context learned during the work process: such as build commands, debugging tips, code style preferences, architectural points, etc., and will automatically recall them in the next session without the user needing to manually record anything. Previously, Claude Code supported files, but those were instructions written by the user to Claude. The newly added feature, on the other hand, is a "notebook" maintained by Claude itself. Users can also directly tell Claude, "Remember we use pnpm instead of npm," and it will write it into the memory file. Technically, the memory for each project is stored in the local directory ~/.claude/projects/, and the first 200 lines of the memory are automatically loaded when the session starts, with more detailed content read as needed. This feature is enabled by default and can be turned off via the /memory command or configuration file. This feature addresses the pain point of developers repeatedly explaining project backgrounds to AI: the longer you use it, the more familiar Claude becomes with the project. Related documentation: