The Gatekeeper Knows Enough
Fikresilase Wondmeneh Abebayew
TL;DR
The paper tackles state desynchronization and context-management bottlenecks in autonomous LLM agents. It introduces the Gatekeeper Protocol, built around a central System State-Context Representation (SCR) denoted by $\mathcal{L}$, which serves as a latent map, authoritative state, and declarative action interface, with deterministic state transitions via $\Phi$. The agent reasons on a low-fidelity latent map and on-demand retrieves high-fidelity context through a unified JSON ledger, achieving improved token efficiency and reliability. Empirical evaluation across three programming tasks and seven LLMs shows that Sage, implementing Gatekeeper, attains higher task completion (e.g., $73\%$) and substantially fewer grounding errors and tokens than baselines, supporting the claim that architectural protocol design, not merely model size, drives robust performance. The work lays a domain-agnostic foundation for grounded, trustworthy AI agents in structured knowledge domains, with future directions including hierarchical latent maps and a universal Gatekeeper specification.
Abstract
Large Language Models (LLMs) are increasingly deployed as autonomous agents, yet their practical utility is fundamentally constrained by a limited context window and state desynchronization resulting from the LLMs' stateless nature and inefficient context management. These limitations lead to unreliable output, unpredictable behavior, and inefficient resource usage, particularly when interacting with large, structured, and sensitive knowledge systems such as codebases and documents. To address these challenges, we introduce the Gatekeeper Protocol, a novel, domain-agnostic framework that governs agent-system interactions. Our protocol mandates that the agent first operate and reason on a minimalist, low-fidelity "latent state" representation of the system to strategically request high-fidelity context on demand. All interactions are mediated through a unified JSON format that serves as a declarative, state-synchronized protocol, ensuring the agent's model of the system remains verifiably grounded in the system's reality. We demonstrate the efficacy of this protocol with Sage, a reference implementation of the Gatekeeper Protocol for software development. Our results show that this approach significantly increases agent reliability, improves computational efficiency by minimizing token consumption, and enables scalable interaction with complex systems, creating a foundational methodology for building more robust, predictable, and grounded AI agents for any structured knowledge domain.