Скачать или смотреть Multi-Agent Multi-User Collab thru MCP: Enabling Asynchronous Human-Agent Agent-Human Workflows

[Session] Multi-Agent Multi-User Collab thru MCP: Enabling Asynchronous Human-Agent Agent-Human Workflows
🎤 Muneeb Arshad, Software Engineer - TELUS
🎤 Andrew Culberson, Senior Software Engineer - TELUS

The Model Context Protocol (MCP) has demonstrated significant value in establishing robust connections between AI agents and various data sources, development tools, and external systems. As the field evolves, there is growing excitement around the potential for direct agent-to-agent communication and collaboration, opening new possibilities for distributed AI workflows.

This presentation will explore the emerging landscape of Multi-Agent Multi-User Workspaces (MAMUWs), examining both the technical foundations and practical applications of these collaborative environments. We will demonstrate how MCP serves as the critical infrastructure layer that enables seamless communication between human users, AI agents, and interconnected agent networks, with particular emphasis on leveraging underutilized protocol features that unlock advanced collaboration patterns.

The session will feature a live demonstration of a production-ready application that facilitates asynchronous collaboration among multiple developers through their individual coding agents. This implementation showcases innovative use of MCP's sampling capabilities to enable agents to request and evaluate multiple solution approaches from peer agents, and elicitation features to dynamically gather context and preferences from human collaborators during the workflow process.

Key topics will include:Technical architecture patterns for multi-agent systems using MCP's full feature set
Practical implementation of MCP sampling for agent-to-agent solution exploration and consensus building
Leveraging elicitation mechanisms to create adaptive workflows that respond to human input and changing requirements
Design principles for human-agent-agent-human collaboration workflows that maximize both AI capabilities and human oversight
Real-world use cases beyond software development, including content creation, data analysis, and project managementChallenges and solutions for maintaining context, security, and reliability in distributed agent networks
Performance optimization strategies when using advanced MCP features at scale
The demonstration will specifically highlight how sampling enables agents to collaboratively explore solution spaces and reach consensus on optimal approaches, while elicitation allows the system to dynamically adapt to user preferences and project constraints without requiring explicit configuration.

Attendees will gain practical insights into implementing sophisticated multi-agent systems that fully utilize MCP's advanced capabilities, understanding how features like sampling and elicitation can transform simple agent interactions into intelligent, adaptive collaboration platforms.

✨ MCP Developers Summit EU 2025
📅 2 Oct 2025
📍 London, UK