Yoke – Build Your Agent over Claude Code and Codex (github.com)

A new Python SDK called Yoke has been announced, designed to facilitate the creation of AI agents using Claude Code and Codex. This SDK allows developers to define agent behaviors, goals, and workflows, and supports both platforms within a unified framework, making it easier to build and manage complex AI systems. By streamlining the integration process, Yoke enables users to leverage existing skills and subagents without needing separate API keys, thus simplifying the development workflow. The significance of Yoke lies in its ability to unify capabilities across two of the leading AI platforms—Claude and Codex—while maintaining their unique features. Developers can create agents that operate seamlessly between these environments, allowing for enhanced collaboration and efficiency. Key features include the ability to define workflows with dependencies, manage state goals, and implement real-time event observation during executions. This flexibility empowers researchers and engineers in the AI/ML community to construct sophisticated agents and workflows more intuitively, potentially accelerating the development of advanced AI applications. As Yoke is still in early alpha, its API may evolve, promising future enhancements and broader functionality.