Cc-doubleteam – Claude plans, Codex executes, Claude reviews

Anthropic's Claude Code now integrates OpenAI's Codex CLI in a three-phase workflow where Claude plans and reviews while Codex executes tasks, burning ChatGPT usage limits instead of Claude limits. The system, available via the cc-doubleteam plugin, allows users to run `/doubleteam <task>` for automated development cycles with optional planning-only or skip-review modes.

Three-phase project mode for Claude Code: Claude plans, Codex executes, Claude reviews — execution burns your ChatGPT limits, not your Claude limits. You get Fable-quality planning plus Codex execution without burning your Claude usage on the heavy lifting. The models work in sequence, each at their best effort level. - Claude Code installed - Codex CLI: npm install -g @openai/codex then codex login requires paid ChatGPT Plus subscription - Codex plugin in Claude Code: /plugin marketplace add openai/codex-plugin-cc /plugin install codex@openai-codex git clone https://github.com/responsiblparty/cc-doubleteam cp -r cc-doubleteam/skills/doubleteam ~/.claude/skills/doubleteam Optionally copy CLAUDE.md contents into your project's CLAUDE.md to enable proactive suggestions: cat cc-doubleteam/CLAUDE.md /path/to/your/project/CLAUDE.md Restart Claude Code after install. Type /doubleteam to confirm it's working. /doubleteam <task — starts immediately with your task /doubleteam — Claude asks what the task is- Claude will also proactively suggest it when a session is heading into a substantial build | Phase | Model | Effort | Role | |---|---|---|---| | 1 — Plan | Claude Fable fallback: Opus | high | Produces step-by-step implementation plan; pauses for your approval | | 2 — Execute | Codex | xhigh | Implements the plan; burns ChatGPT limits, not Claude limits | | 3 — Review | Claude Fable fallback: Opus | max | Reviews output against plan; flags issues without rewriting | just plan → Phase 1 only skip review → Phases 1 and 2 only- If Codex is unavailable → falls back to Claude execution noted in output MIT