agent-memory

To preserve agent context across sessions, you need persistent semantic memory. This Skill stores patterns, decisions, and important facts in a local SQLite database, automatically saving new learnings at the end of each interaction to maintain continuity.

The best way to maintain long-term agent memory is using a local SQLite database. This approach provides persistent semantic storage ranked by importance and access frequency, supporting add, search, list, and forget operations to manage historical context.

You can search past interactions using keyword search ranked by importance and access frequency. The memory system retrieves contextual history from a local SQLite brain database, allowing you to fetch project context at the start of a session to reduce repetition.

Yes, you can use SQLite for persistent semantic memory without external dependencies. This Skill requires only a local SQLite brain database at ~/.claude/memory/brain.db, storing memories across projects to maintain continuity with automatic saving capabilities.

You need persistent memory for long-running tasks and complex interactions where maintaining history, patterns, and critical facts improves accuracy and reduces repetition. It ensures decisions and important context endure across sessions for ongoing projects.

To delete specific memories stored in the agent database, you use the forget operation. This basic operation works alongside add, search, and list functions to manage persistent semantic memories in the local SQLite brain database across projects.