When choosing an AI coding agent from the current field (e.g., Claude Code, Cursor, Devin), what is your primary decision criterion?

Options

• Alignment with my preferred workflow (IDE-centric vs. terminal-first).
• Autonomy tolerance and ability to handle multi-step tasks reliably.
• Token cost efficiency and performance on complex, large-scale problems.
• Integration with existing tools and frameworks, like `AGENTS.md` and common APIs.