A comprehensive comparison of terminal-based AI coding agents including Codex CLI, Charm Crush, OpenCode, and Cursor CLI, testing their ability to work with GPT-5 and support essential features like MCP servers and saved prompts. The evaluation reveals significant shortcomings across all tested agents, with most failing basic functionality requirements, while Claude Code remains the only fully functional option despite being limited to Anthropic models.
•33m watch time
1 Comment
Sort: