Learn how to embed the Codex agent using the Codex App Server, a bidirectional JSON-RPC API powering streaming progress, tool use, approvals, and diffs.

OpenAI is a research organization focused on artificial intelligence and machine learning. Readers can learn about  AI research, deep learning models, and AI applications across various domains. With research papers, blog posts, and technical documentation, OpenAI provides  insights and expertise for understanding and advancing the field of artificial intelligence.

OpenAI

OpenAI's Codex App Server is a bidirectional JSON-RPC API that enables different surfaces (web, CLI, IDE extensions, macOS app) to embed the same Codex agent harness. The architecture includes thread lifecycle management, authentication, tool execution, and a protocol built around three primitives: items (atomic I/O units), turns (agent work units), and threads (persistent conversations). The App Server evolved from an internal VS Code integration into a stable platform that supports streaming progress, tool approvals, and diffs. Integration patterns include local apps bundling the binary, web runtime using containers, and the TUI being refactored to use the protocol. The open-source implementation provides JSON Schema generation for building clients in multiple languages.

Unlocking the Codex harness: how we built the App Server