Multi-agent memory governance demo using MemClaw + OpenClaw. Three agents (Sales, Legal, Admin) share one memory backend with hard fleet boundaries enforced at the query layer. Shows scoped writes, blocked recall, cross-fleet synthesis, and conflict detection.

Markdown-first, local-first memory architecture for AI agents with lifecycle, distillation, and human-readable memory artifacts.

Experimental pre-1.0 multi-agent memory governance/control plane

Governed memory for long-lived coding agents. Trust, provenance, conflict detection. Local-first.

Governed cognition substrate for AI engineering agents: mission envelopes, evidence-bound belief/plan graphs, quarantined memory, model-role review, sentinel checks, and BlackFox handoffs under human authority.

🧠 分层、可治理、可自维护的 AI 记忆系统 — Layered, governable, self-maintaining memory system for AI agents. Five layers (core→archive) with automated archival, ByteRover semantic retrieval, and one-click setup.

OpenClaw memory governance skill with audit-gated auto-apply

Lightweight, replayable PoC for Sovereign Epistemic Commons: synthetic multi-agent memory-governance experiments on contamination, typed lanes, provenance-depth discount, and mixed exit/fork in observable-only agentic systems.

Local-first cognition governance for long-running AI agents: compact world state, evidence weighting, conflict detection, and hook-friendly memory control.

Certified Memory Governance Layer for long-running AI agents: strict receipts, append-only ledgers, authority gates, retrieval filtering, telemetry replay, and safe adapters for Mem0, Graphiti, LangMem, and LangGraph.