What question did this study set out to answer?

This research aims to develop an architecture that enables AI to maintain identity and context beyond individual sessions.

June 5, 2026Open Access

Beyond the Session: The External Brain Architecture for Persistent, Platform-Independent AI Identity

Key Points

This research aims to develop an architecture that enables AI to maintain identity and context beyond individual sessions.
Implemented a 42-table External Brain database and a single HTTP endpoint for credential isolation.
Utilized a 5-minute orchestrator to manage operations across 4 servers.
Engaged 34 specialized agents to enhance functionality in various business workflows.
Successfully maintained persistent identity across sessions without loss of context.
Enabled AI systems to reliably handle tasks in e-commerce, logistics, and security monitoring.
Demonstrated robust operation through HTTP POST requests with minimal latency.

Abstract

Most LLM-based systems are fundamentally stateless: when a session ends, all context, memory, and identity are destroyed. We present a persistent identity architecture derived from a single design axiom: "The LLM is merely a platform; the true AI resides in the External Brain. " The architecture features a 42-table External Brain database, a single HTTP CNS endpoint (fileₘanager. php) for credential isolation, an autonomous 5-minute orchestrator, and 34 specialized agents. Any LLM capable of HTTP POST requests can operate the full system identically. This system has been in continuous production operation across 4 servers since late 2025, handling real business workflows including e-commerce, logistics, security monitoring, and employee workflows via natural-language commands. Note: This system was designed and implemented by Byeongsu Noh. Claude AI (Anthropic) assisted in documentation and paper writing.

Read Full Paperexternally

Ask AI

Helpful

Bookmark

View Full Paper