CoDesignAI: An AI-Enabled Multi-Agent, Multi-User System for Collaborative Urban Design at the Conceptual Stage

A research team led by Zhaoxi Zhang has introduced CoDesignAI, a novel web-based platform that leverages large language models (LLMs) to transform the conceptual stage of urban design. The system creates a multi-user, multi-agent environment where human participants (representing residents or stakeholders) collaborate alongside specialized AI agents that act as domain experts. These AI agents provide critical facilitation and professional knowledge, helping to structure discussions, summarize content, and extract shared design intentions from often chaotic public input. The architecture is designed to make citizen engagement in urban planning both more efficient and scalable, addressing a long-standing bottleneck in participatory design.

CoDesignAI's technical core integrates conversational AI with spatial mapping services, enabling the real-time, street-level visualization of design proposals. As users discuss potential interventions—like adding a park or changing traffic flow—AI agents can generate prompts that translate these ideas into concrete visual representations grounded in the actual urban context. The platform supports an iterative workflow where proposals can be revised and refined over multiple rounds, with the entire design process automatically documented. This combination of multi-user interaction, AI mediation, and image-based design grounded in real-world maps represents a significant step toward making urban design a more open and collaborative process, rather than one dominated solely by professional experts.