session-memory

Persistent memory across Claude Cowork sessions is maintained by storing user context, preferences, and decisions in a dedicated CLAUDE-MEMORY.md file. This allows the assistant to load existing context and update it during future interactions without manual re-entry.

Yes, you can save and load assistant preferences by instructing the session to read from or write to a CLAUDE-MEMORY.md file. Use natural language trigger phrases to load stored memory at the start and save updates at the end of a session.

Managing long-running project planning across multiple sessions is best handled by persisting ongoing decisions and project states in a memory file. This ensures context continuity so the assistant remembers previous interactions and adapts to evolving project requirements.

No, you do not need to manually copy context into the memory file. The assistant updates the persistent memory automatically when you use specific natural language trigger phrases to save the current session's context and decisions.

Persistent memory works without external dependencies by using a markdown file as the storage mechanism. The setup requires a SKILL.md frontmatter with a name and description, while scripts and other references are optional for the core memory functionality.

You should not use a persistent memory file for session context if your workflow involves sensitive data that should not be stored in plain text, as the memory is saved directly into a standard markdown file in your user folder without encryption.