Composable Contracts for Multi-Agent Coordination

Abstract

Cooperative AI faces information asymmetry problems, particularly when agents built on different systems interact with each other. Decentralization of data, contracts, and finance offer solutions to these challenges. We propose a programmable contract framework that modularly composes tasks and scales interdependent sequences in distributed flows of tasks and agents. These contracts mitigate informational friction, align agent actions and payoffs, and ensure credible commitments. Additionally, we explore how market mechanisms can further facilitate contract composition and workflow efficiency.