This package provides an authentication plugin for the OpenCode AI platform, enabling users to leverage their existing ChatGPT Plus/Pro subscriptions with the OpenAI Codex backend instead of consuming API credits. The current stable version is 4.4.0, and the project demonstrates a very active release cadence with multiple minor and patch releases in quick succession, indicating ongoing development and maintenance. Key differentiators include a simplified one-command installer (`npx opencode-openai-codex-auth@latest`), comprehensive support for various GPT-5.x and Codex model families (including GPT-5.2 Codex), and an official OAuth flow for seamless integration. It handles multimodal input and automatically manages token refresh, aiming to provide a 'one config, every model' philosophy for OpenCode users. It is designed primarily for personal development use, with a warning against production or multi-user applications.
npm install opencode-openai-codex-authVerified import paths — ran on the pinned version, not inferred.
This quickstart demonstrates the one-command installation and subsequent authentication and model usage with the OpenCode CLI.
Update all configurations to use GPT-5.1 model identifiers. Refer to `config/full-opencode.json` for the officially supported configuration structure.
Switch all references from `gpt-5.0` to `gpt-5.1` in your OpenCode commands and configurations, specifically updating model presets.
Avoid setting `reasoning_effort: "none"` for Codex variants. If faster responses are desired, use other reasoning effort levels or non-Codex GPT models that support 'none'.
For production or multi-user applications, utilize the official OpenAI Platform API directly instead of this plugin to ensure proper scalability and credential management.
Upgrade to version 4.0.2 or newer, which contains fixes for agent creation failing and correct handling of SSE/JSON responses.
Upgrade to version 4.0.1 or newer. This version fixes the compaction context loss by only filtering orphaned outputs (without matching calls), preserving matched pairs for context.
Upgrade to version 4.0.1 or newer, which includes a fallback mechanism to handle GitHub API rate limiting.