Codex Knowledge Base
Tessl Skill Evaluation Framework: Treating Agent Skills as Production Software
You have written a skill for your coding agent. It looks right. It seems to work when you try it.
Plan Mode Mechanics: Enter vs Tab, Syntax Highlighting and Inline Editing
Codex CLI's terminal interface has evolved well beyond a simple prompt-and-response loop. Between plan mode, steer mode, and a rich set of TUI enhancements.
Migrating to Codex CLI from ChatGPT: From Chat to Agentic Workflows
If you have spent the last two years refining ChatGPT prompts — crafting custom instructions, building up conversation histories, and learning to coax code.
Installing Codex CLI: Package Managers, Shell Completions and First-Run Setup
Getting Codex CLI running on your machine takes under five minutes — but the installation path you choose and the post-install configuration you apply can.
Gas Town: Steve Yegge's Multi-Agent Factory and What It Means for Codex CLI
In January 2026, Steve Yegge open-sourced Gas Town — a Go-based multi-agent workspace manager that orchestrates 20–30 parallel Claude Code instances under a.
The Four-Surface Architecture: CLI, Desktop, IDE Extension and Cloud as One System
OpenAI ships Codex as a single agent with four points of entry. The CLI, Desktop app, IDE extension and Cloud surface all share the same configuration.
Cross-Surface Session Sync: Resuming Codex Sessions Across CLI, Desktop and Cloud
Codex is no longer a single-surface tool. With the CLI, Desktop app (macOS and Windows), IDE extension and Cloud runtime all sharing a unified backend.
Cross-Platform Agent Portability: One SKILL.md for Every Runtime
The AI coding agent landscape has fragmented into over a dozen serious contenders.
Codex TypeScript SDK: Streaming, Multimodal Inputs and Per-Thread Configuration
The Codex TypeScript SDK (@openai/codex-sdk) transforms Codex CLI from an interactive terminal tool into an embeddable agent runtime .
Inside the Codex Sandbox: Platform-Specific Implementation on macOS, Linux and Windows
Codex CLI's sandbox is not a single mechanism — it is three distinct OS-native enforcement layers unified behind one policy abstraction. Understanding what.