managing-memory

You can manage memory blocks by inspecting, editing, auditing, and restructuring them directly from the local filesystem or via the Letta API for subagents, keeping memory architecture in sync and preventing bloat.

The Letta API manages subagent memory blocks by handling API calls to coordinate memory at scale, allowing you to update and synchronize block values across scout, coder, and other subagents.

Auditing agent memory utilization surfaces block usage and health metrics to optimize the memory footprint, prevent bloat, and identify when memory architecture requires cleanup or reevaluation.

Yes, you need the letta_client dependency to execute Letta API calls for cross-agent memory operations, which is required for managing memory blocks across scout, coder, and other subagents at scale.

You should invoke the memory agent to trigger major restructures when the memory architecture needs reevaluation or cleanup, handling complex block restructuring that goes beyond basic editing and auditing.

Yes, you can edit and read memory blocks directly from the local filesystem using the preferred Memory Filesystem path, keeping memory block labels and values in sync without requiring API calls.