para-memory-files

You can persist agent memory using a file-based three-layer structure under $AGENT_HOME, storing durable context as atomic YAML facts, markdown daily notes, and instruction-style tacit memory to retain knowledge across sessions.

The PARA method organizes persistent agent knowledge into structured folders, storing durable entity-scoped facts as atomic YAML files and keeping summaries refreshed through weekly synthesis for reliable file-based recall.

Record daily notes by appending raw timeline information into date-stamped markdown files, creating a continuous and traceable log of when events occurred for reliable session recall.

Yes, the system maintains instruction-style tacit memory that captures how the user operates, automatically updating these stored patterns as new behaviors are learned during interactions.

Manage shared plans by storing them in a designated plans directory and retrieving directly from on-disk sources of truth, avoiding scattered ad-hoc notes for better file-first recall.

Entity fact updates use non-destructive superseding semantics, ensuring that writing new facts into items.yaml preserves the historical context and prevents data loss during knowledge modifications.