Which Tool When: A Comparison
A cheat sheet to match the tool to the job.
Here's the whole landscape on one page.
By interface
| Tool | Interface | Best for | | --- | --- | --- | | Claude Code | Terminal | Deep repo work, power users | | Codex CLI | Terminal | OpenAI users, sandboxed CLI | | Gemini CLI | Terminal | Huge context, budget | | Aider | Terminal | Git-native, scriptable | | Cursor | Editor (VS Code fork) | Editor-first workflow | | Windsurf | Editor | Visual agent reasoning | | Copilot | Editor + browser | GitHub-centric teams | | Claude Desktop | Desktop chat | MCP experimentation | | Devin / OpenHands | Cloud agent | Full task delegation |
By situation
- Quick inline edits while typing → Cursor Tab, Copilot completion.
- A tricky multi-file change you want to watch → Claude Code or Cursor Agent.
- Delegate a whole ticket and review a PR later → Codex cloud, Devin, OpenHands.
- No budget / offline / private code → Ollama + Aider or Continue.
- Learn how agents think → Windsurf Flows, or this course's Playground.
A personal starter stack
- Editor: Cursor (or VS Code + Copilot).
- Terminal: Claude Code.
- Local: Ollama + one open coding model.
- MCP: the same servers across all of them.
Don't optimize for the "best" tool. Optimize for the one you'll actually use every day — and let MCP carry your tools between them.