在 token 預算下運作的 AI 團隊:Grok Bot、Claude Code 與 Codex
My AI Team on a Token Budget: Grok Bot, Claude Code, and Codex
作者在 Grok Bot 最低階的付費方案中組建小型團隊,並讓團隊使用 Claude Code 和 Codex 的入門付費方案。Grok Bot 負責簡短指示和協調,較繁重的工作則交由 Claude Code 或 Codex;各 Agent 在共用工作區記錄進度和交接事項。作者表示,這套安排讓他完成更多工作,也更有效運用 token 預算。
Hi everyone! 👋
I’ve been trying a setup for my side projects, and I’m curious whether anyone else is doing something similar.
I set up a small team in Grok Bot on its lowest paid plan and gave the team access to Claude Code and Codex. I use entry-level paid plans for those tools too.
I gave every agent two rules:
- Use Claude Code or Codex for the heavy lifting. My Grok Bot token budget is tighter, so I save it for short instructions and coordination. When a task involves digging through a project, making substantial changes, or reviewing a lot of code, the team hands it off to Claude Code or Codex.
- Leave progress and handoffs in the shared workspace. The agents use tools already available on their machine, like spreadsheets and working folders, to keep track of what they did, where the files are, and what’s next. The next agent can pick up from there and keep bot-to-bot chat to a minimum.
So far, I’ve been getting more done, and it feels like I’m making better use of my token budget. A small thing that helps is having the next agent open a file with the test results instead of asking another agent to explain the whole task again.
Is anyone else running a setup like this? I’d love to hear what you delegate to Claude Code or Codex, how your agents hand off work, and whether you’ve noticed a difference in how quickly you hit your usage limits.
來源:DEV Community · dev.to