🧠

Memory System

Claude Code's built in persistent memory that automatically stores user preferences, feedback, project context, and references across sessions. Learns from your corrections and remembers your patterns.

Data Source·2 sections·1 min read

productivityworkflow

Install Prompt

Paste this into Claude Code to set it up:

Claude Code's memory system works automatically. To see your memories, check ~/.claude/projects/*/memory/ directories. You can ask Claude to 'remember' something and it will save it, or 'forget' something to remove it. Memory types include: user (who you are), feedback (how to work), project (ongoing context), and reference (where to find things).

01What It Does

Automatic cross session learning

Claude Code maintains a persistent memory file for each project at ~/.claude/projects/<project>/memory/MEMORY.md. This file is populated automatically as Claude notices patterns, receives corrections, and accumulates context across sessions. You never need to configure it. It starts working from your first interaction.

User memories: Who you are, your background, your role, your communication preferences. Claude remembers these so you never have to reintroduce yourself.

Feedback memories: Corrections and preferences about how Claude should work. When you say "never use dashes in your writing" or "always ask before committing," Claude encodes that as a persistent rule.

Project memories: Ongoing context about what you are building, current state of work, decisions made, and progress across sessions. The thread that connects separate conversations.

Reference memories: Where to find things, which files matter, key paths and URLs. The map that helps Claude navigate your world efficiently.

02How It Works

Save

Claude observes patterns, receives feedback, and encounters important context during your session. These get written to the MEMORY.md file as structured entries with clear categorization.

Index

Each memory is organized by type (user, feedback, project, reference) with descriptive headings. This structure makes retrieval fast and prevents memories from becoming an unreadable dump of notes.

Recall

At the start of every session, Claude loads the memory file and applies what it knows. Your preferences shape its behavior, project context informs its suggestions, and references help it navigate your codebase.

Update

Memories evolve. When you correct something, Claude updates the relevant entry. When context changes, outdated memories get replaced. You can also edit MEMORY.md directly to add, modify, or remove any entry.

You are always in control. Read your MEMORY.md files anytime to see what Claude has learned. Edit them directly to correct mistakes, add important context, or remove entries that are no longer relevant. The memory system is transparent by design.