Store, recall, and connect knowledge in a single SQLite file. FTS5 search, hybrid embeddings, facts extraction, and knowledge graph — all offline, zero external dependencies.
Extract facts, link entities, and traverse relationships. Find paths between concepts your agent has learned.
FTS5 lexical search + semantic embeddings with configurable reranking. Best of both worlds.
50% storage reduction with F32→F16 conversion. Embeddings stay fast and compact.
15 tools for Claude Desktop integration. Store, recall, search facts, explore the graph.
Plugin automatically stores conversations with attribution. Entity + session grouping.
Auto-extract structured facts from messages (confidence ≥ 0.6). Build knowledge hands-free.
# Core library
npm install @akashabot/openclaw-memory-offline-core
# CLI tool
npm install -g @akashabot/openclaw-mem
# MCP Server (optional, for Claude Desktop)
npm install -g @akashabot/openclaw-memory-mcp-server
# OpenClaw plugin (bridge to core)
# https://github.com/AkashaBot/openclaw-memory-offline-sqlite-plugin
# Initialize and test
openclaw-mem init
openclaw-mem remember "OpenClaw runs offline memory"
openclaw-mem search "offline" --hybridplugins:
load:
paths:
- ./plugins/memory-offline-sqlite
slots:
memory: memory-offline-sqlite
entries:
memory-offline-sqlite:
enabled: true
config:
dbPath: ~/.openclaw/memory/offline.sqlite
autoRecall: true
autoCapture: true
mode: hybrid
provider: ollama
embeddingModel: bge-m3