
from smart-memory22
A transcript-first cognitive memory system for agents with revision-aware retrieval and bounded prompt composition.
Smart Memory implements a sophisticated local cognitive OS for AI agents. Unlike simple vector stores, it uses a transcript-first approach where memories are derived from actual conversation logs. It supports typed long-term memory (preferences, identity, task state), evidence-backed revisions, and explicitly managed memory lanes.
Activate this skill when an agent requires deep continuity across long sessions, needs to remember specific user preferences reliably, or must maintain a complex state of an ongoing task without losing context in the prompt window.
server.py, cognitive_memory_system.py).memory_search, memory_commit, and memory_insights to maintain agent continuity.Specifically integrated for OpenClaw, but compatible with any agent that can interact with a FastAPI backend and Node.js adapter.
This skill has not been reviewed by our automated audit pipeline yet.