Agentic LLM systems that generate code through multi-turn tool use face a context problem: each session starts from zero, discarding configuration choices, domain constraints, data schemas, and tool-use patterns that made previous sessions productive. Naively persisting entire conversation histories is token-inefficient and counterproductive, as irrelevant context degrades generation quality. The authors introduce shared selective persistent memory, a memory architecture that identifies and retains four categories of reusable context—task specifications, data schemas, tool configurations, and output constraints—while discarding session-specific reasoning traces. This memory is shared across users with role-based access control, enabling collaborative reuse without redundant specification.

The architecture is implemented in a deployed collaborative workspace platform where LLM agents produce, edit, and maintain git-versioned artifacts such as interactive dashboards, structured reports, and data-driven documents. Data sources include CSV uploads, SQL, REST APIs, and MCP servers. Git-backed versioning with draft isolation allows users to explore modifications risk-free and restore to any prior state without re-invoking the model. A zero-token data refresh mechanism decouples generated programs from runtime data, enabling artifact reuse without re-invocation.

Reported results across three enterprise deployment scenarios show that shared selective persistent memory achieves 96% task completion, compared to 79% without memory and 71% with full history. The zero-token data refresh mechanism eliminates LLM re-invocation for recurring data updates, yielding a 14× task time reduction. Summary-driven generation reduces per-invocation token cost by 97× versus raw data injection. A replication on four public datasets confirms generalizability, with zero-token refresh succeeding in 12/12 trials. Notably, naive full-history persistence actively degrades task completion by biasing the agent with stale reasoning traces, while selective memory outperforms both extremes.

The results indicate that selective persistence of structured context types is more effective than both no memory and full history. The authors suggest that shared memory facilitates collaboration by transferring accumulated context across users with role-based access, reducing redundant specification. The zero-token refresh mechanism is a practical solution for recurring data updates. The source also includes related readings on tool-use evaluation and false-trigger mitigation in speech interfaces, but these are separate and not part of the main architecture.