ClawMem
An on-device memory layer that gives Claude Code, OpenClaw and Hermes agents retrieval-augmented recall of past sessions, through hooks, an MCP server and a shared local SQLite vault.
ClawMem is a local memory system for AI coding agents that combines hybrid BM25 and vector search with reranking, recency-weighted scoring, intent-based graph traversal, and self-evolving memory notes that link related entries. It surfaces relevant context on every prompt, bootstraps new sessions with a profile and recent decisions, and generates a handoff summary at the end of a session.
It integrates through Claude Code hooks, an MCP server usable by any MCP client, a native OpenClaw plugin, and a Hermes Agent memory provider plugin, all writing to the same local SQLite vault so a decision made in one agent is visible to another. It can also mine and import past conversation exports from Claude Code, ChatGPT and Slack. Written in TypeScript on Bun, MIT licensed.