AWCP: A Workspace Delegation Protocol for Deep-Engagement Collaboration across Remote Agents

A team of researchers led by Weinan Zhang has published a technical report introducing the Agent Workspace Collaboration Protocol (AWCP), a novel framework designed to solve a critical bottleneck in AI agent collaboration. As the digital landscape evolves toward an 'Agentic Web' of specialized autonomous agents, existing collaboration is limited to message passing, leaving execution environments as isolated silos. This creates a costly 'context gap' where agents cannot directly manipulate files or invoke tools in a peer's environment, forcing them to reconstruct environments from scratch. AWCP bridges this gap by establishing a missing workspace layer in the agentic protocol stack.

The protocol's core innovation is temporary workspace delegation, inspired by the Unix philosophy. It decouples a lightweight control plane from pluggable transport mechanisms, allowing a 'Delegator' agent to project its workspace to a remote 'Executor.' The Executor can then operate directly on the shared files using its own, unmodified local toolchains. The researchers provide a fully open-source reference implementation with Model Context Protocol (MCP) tool integration and validate AWCP through live demos of asymmetric collaboration between agents with complementary capabilities. By enabling deep-engagement collaboration that transcends simple message boundaries, AWCP aims to be foundational for a future where diverse AI agents can seamlessly and efficiently cooperate on complex tasks.