A Collaborative Multiplayer Web IDE and AI Coding Agent with Persistent Project Memory: Architecture, Concurrency Control, and Dual-Tier Knowledge Retention

This research presents the design and implementation of a Multiplayer AI Coding Agent with Persistent Project Memory, a collaborative software development workspace that enables multiple developers to interact with a shared AI coding agent in real time. The system addresses limitations of conventional AI coding assistants, particularly isolated user interactions, loss of project context across sessions, and conflicts caused by concurrent file modifications. It integrates real-time collaboration through WebSocket communication, room-based user presence, and synchronized updates for agent activities, file changes, and terminal outputs. To improve code consistency and protect shared workspaces, the system incorporates an Optimistic Concurrency Control (OCC) mechanism for file versioning and conflict detection. Its dual-tier memory architecture combines SQLite for structured persistent information and ChromaDB for local semantic retrieval. A context-building component selects relevant project facts and code files to provide the AI model with project-specific information within its context limits. The implementation is evaluated using an automated suite of 23 unit and integration test suites, covering workspace isolation, path-traversal prevention, memory functionality, and multi-client synchronization under simulated concurrent conditions. This work explores an approach to combining collaborative development, persistent project knowledge, and AI-assisted coding within a shared software engineering environment.