Show HN: Claude Code Workflows for GPT/Codex (github.com)

A new multi-agent workflow runtime has been announced, leveraging the Codex App Server to enhance the capabilities of GPT/Codex. This workflow allows developers to structure their JavaScript code with control flow elements such as loops and retries, while also introducing key improvements like resumability and validated structured output. The system stores completed calls in a durable journal, thus enabling the replay of previous calls without incurring additional costs. Developers can ensure output conforms to a specified JSON schema, automatically validating responses and retrying invalid ones. This release holds substantial significance for the AI/ML community as it streamlines the development and debugging of AI workflows, making it easier to manage complex tasks that involve multiple agent interactions. The introduction of independent verification processes allows for improved fault tolerance; when a call fails, it resolves to a null value rather than aborting the entire workflow, thereby enhancing robustness. Built to require Bun 1.3 or newer, with a reliable CLI for execution, this tool promises to transform how developers interact with Codex servers and structure AI-driven workflows efficiently.