memory-manager

To store and retrieve agent long-term memory across sessions, use a deterministic CLI to search past diary entries and MEMORY.md snippets. This coordinates layered storage across MEMORY.md, diary/, and memory/ to balance context length with depth, enabling recall of key decisions and milestones.

Managing agent memory for post-mortems and decision reviews requires extracting relevant past events and preferences from a long-term memory system. The memory CLI provides fast retrieval and snippet extraction from MEMORY.md and diary/ to inform current planning approaches.

Yes, you can use a memory CLI to search past diary entries in your current workspace. Running the agent/memory_cli.py tool enables fast search and get operations to retrieve relevant historical notes, preferences, and milestones from your layered memory storage.

Layered memory storage balances context length with depth by coordinating MEMORY.md, diary/, and memory/ to preserve critical context. This ensures that agents can recall complex past decisions without overloading the active context window during planning or retrospectives.

Yes, the memory management system supports self-improvement logging. Alongside storing and retrieving an agent's long-term memory, it enables logging for self-improvement to help review past events and guide future project planning approaches.

You need long-term memory for your AI agent before tackling a complex task or planning a project retrospectively. It allows you to pull relevant past decisions, preferences, and milestones to guide the current approach and prevent repeating historical mistakes.