Player not loading? Watch on YouTube
This comparison examines Claude Code, Codex and OpenCode as terminal coding assistants. The speaker prefers Claude Code's permission prompts and OpenCode's terminal interface. His demonstrations show configurable approval settings, so the preferences reflect the setups he presents rather than a universal ranking.
Model access is a central distinction. The speaker says OpenCode supports local models and multiple providers, including OpenAI subscription access and Anthropic API keys. He says Claude subscription access requires Claude Code. For readers considering a local LLM workflow, OpenCode's model choice is relevant, though the video does not demonstrate a local installation or hardware requirements.
The tooling discussion covers Claude Code's rewind, voice input and plugin marketplace, plus OpenCode's session management and undo. The speaker criticizes Codex for lacking undo in the version he uses. OpenCode's redo restores conversation history but does not restore the deleted game file in his demonstration.
A shared Sudoku prompt requests a Python backend and an HTML, JavaScript and CSS frontend. The speaker finds roughly similar results from Claude Code, Codex and OpenCode using GPT, while several other model outputs have bugs or setup failures. This single task is not a general benchmark. He favors OpenCode's open source approach but uses Claude Code for most engineering work, switching to OpenCode with GPT when he reaches usage limits.
