18. Agent memory — what agents remember, and how you keep it
This chapter has moved into pages, one per thing you want to do.
| What you want to do | Page |
|---|---|
| Choose between instruction files, the agent’s own memory and Agent Fleet memory | Choose which memory your agents use |
| Read, correct, pin and forget what agents remember | Read and correct agent memory |
| Check who changed what and when, and revert it | Check memory changes |
| Let sessions use memory, or stop them | Set how memory is used |
| Check memories for staleness | Find stale memories |
| Have a model draft fixes to memory | Use memory tidy-up |
| Import or write back the memory of Claude Code and others | Import and export memory |
| Roll an agent’s own memory back with snapshots | Restore memory from a snapshot |
| Use Agent Fleet memory from an external agent | Use memory from an external agent |
Agent memory
Section titled “Agent memory”The Memory view, Changes and Settings are now described in the pages that start at Read and correct agent memory.
Import & export
Section titled “Import & export”Importing memory from Claude Code, muse or a bundle, writing it back to Claude Code and moving the memories of a project that no longer exists are described in Import and export memory.
Memory snapshots
Section titled “Memory snapshots”The version history of the memory an agent keeps in its own files is described in Restore memory from a snapshot.
Using Agent Fleet memory from an external agent
Section titled “Using Agent Fleet memory from an external agent”Using Agent Fleet memory from your own Claude Code, Claude Desktop or claude.ai, and bringing your own Claude Code’s memory in, are described in Use memory from an external agent.