Keyoku
Keyoku is a persistent memory system for AI agents that solves the stateless session problem — by default, agents forget everything between conversations.
About
Keyoku is a persistent memory system for AI agents that solves the stateless session problem — by default, agents forget everything between conversations. It captures facts in real time during each session, builds a knowledge graph of entities and their relationships, and automatically injects relevant context before each agent response. The storage is local-first using SQLite and an HNSW vector database, so your data stays on your machine. A standout feature is its heartbeat system: 12 SQL-driven signal checks that run autonomously to surface deadlines, sentiment shifts, and recurring themes without waiting to be asked.
Developers building OpenClaw agents that need to maintain context across multiple sessions, track evolving information over time, or autonomously monitor and act on patterns in what they've previously learned.
Pros & Cons
Pros
- check Local-first storage means no data leaves your machine — a meaningful advantage for privacy-conscious deployments
- check The proactive heartbeat can operate at three autonomy levels (observe, suggest, or act), letting you dial in how much initiative the memory layer takes
- check One-command setup via `npx @keyoku/openclaw init` keeps the barrier to entry low for a system with meaningful technical depth
Cons
- close TypeScript/Go stack means developers outside those ecosystems face a steeper integration path
- close The knowledge graph and vector search features add real power but also real complexity — lightweight use cases may not justify the setup
More Memory
Other tools in the same category.
