AOMS - Always-On Memory Service

AOMS is a production memory layer for AI agents. It runs as a local service (FastAPI + systemd), stores episodic/semantic/procedural memory in JSONL, and adds a Cortex engine for progressive disclosure (L0/L1/L2) to slash token costs by 95-99%. Memories are weighted over time so useful experiences surface first and noise decays. Drop it into any agent stack in minutes via REST or CLI. Works with OpenClaw out of the box.

Features:

• 4-tier memory (working/episodic/semantic/procedural)
• Progressive disclosure (L0/L1/L2) - 95% token reduction
• Weighted retrieval with reinforcement learning
• REST API + CLI + Docker
• OpenClaw native integration

GitHub: https://github.com/dhawalc/cortex-mem