Governed Shared Memory for Multi-Agent LLM Systems

2026-06-23

AI Summary

This study addresses four critical failure modes in shared memory systems for multi-agent large language model (LLM) fleets: unauthorized leakage, information staleness, persistent contradictions, and provenance breakdown. To tackle these challenges, the work formally introduces the “fleet memory” problem and proposes a governed shared memory mechanism grounded in system-level primitives—scoped retrieval, temporal coverage, provenance tracing, and policy-based control. Implemented via MemClaw, a multi-tenant memory service, and evaluated using ArgusFleet, a reproducible testing framework, the system enables fine-grained scope enforcement, asynchronous contradiction detection, and synchronized write-gating coordination. Experiments demonstrate that the approach achieves 100% accurate reconstruction of four-layer-deep provenance chains within sub-second latency, ensures zero cross-fleet leakage, and guarantees immediate write visibility in strong-write mode with a single round-trip search.

Abstract

Multi-agent LLM environments require robust mechanisms for shared knowledge management. This paper formalizes the fleet-memory problem and identifies four foundational failure modes: unauthorized leakage, stale propagation, contradiction persistence, and provenance collapse. To address these, we define explicit systems-level primitives: scoped retrieval, temporal supersession, provenance tracking, and policy-governed memory propagation. These primitives are implemented in MemClaw, a production multi-tenant memory service, and evaluated via ArgusFleet, a reproducible harness testing four governance dimensions. Rather than a baseline comparison, this study measures a live production service, emphasizing real-world architectural insights and negative results. Key Evaluation Results Provenance: Successfully reconstructed 100% of depth-four derivation chains with correct writer identity at sub-second per-hop latency. Propagation: Demonstrated high intra-fleet visibility with zero cross-fleet leakage. Under strong write mode, write-to-visible latency was optimized to a single search round-trip. Production Architectural Issues Discovered Asymmetric Scope Enforcement: Tenant isolation held, but sub-tenant scope was initially bypassed on direct GET-by-id requests for agent-scoped credentials (disclosed and remediated during the study). Pipeline Ordering Conflict: While contradiction supersession works for admitted writes, a synchronous near-duplicate gate can prematurely reject contradictory writes before the asynchronous contradiction detector can evaluate them. Conclusion: Long-context retrieval alone is insufficient for production multi-agent memory. Governed shared memory demands explicit systems-level abstractions, and live evaluation is vital to expose enforcement and pipeline-ordering failures missed by design-only treatments.

Problem

• multi-agent LLM
• shared memory
• memory governance
• provenance
• contradiction persistence

Innovation

• governed shared memory
• multi-agent LLM systems
• fleet-memory problem
• systems-level primitives
• provenance tracking