Stop Your AI Agent Forgetting Everything Between Sessions
You know that feeling — you start a new session and your assistant has no memory of yesterday's architecture decisions, your project conventions, or that tricky bug you spent hours debugging?
I maintain an open-source memory sidecar that solves this. v3.5.1 just shipped with some significant upgrades.
What it does:
- Persists context across sessions — hot memory for instant recall, warm memory with semantic search, cold memory for long-term archival, plus curated knowledge notes
- Agent-agnostic — works with Hermes, Claude Code, Codex, Cursor, any agent that has a writable home directory and state.db
- 28 scripts deployed, including session archiving, governance rebuild, alert handling, and dashboard serving
New in v3.5.1:
- Webhook receivers (Telegram, Slack, Feishu/Lark, DingTalk) for action-needed alerts
- OpenMetrics export + Grafana dashboard for SLO tracking
- Synthetic recall benchmarking for regression detection
- Dual-profile isolation soak for multi-agent deployments
MIT licensed. GitHub: https://github.com/mage0535/hermes-memory-installer
Top comments (0)