经 AI Skill Hub 精选评估,Codex-YOLO 获评「推荐使用」。这款MCP工具在功能完整性、社区活跃度和易用性方面表现出色,AI 评分 7.5 分,适合有一定技术背景的用户使用。
Codex-YOLO 是一款遵循 MCP(Model Context Protocol)标准协议的 AI 工具扩展。通过 MCP 协议,它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务,实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用,都可以通过自然语言在 AI 对话中直接触发,极大提升生产效率。
Codex-YOLO 是一款遵循 MCP(Model Context Protocol)标准协议的 AI 工具扩展。通过 MCP 协议,它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务,实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用,都可以通过自然语言在 AI 对话中直接触发,极大提升生产效率。
# 方式一:通过 Claude Code CLI 一键安装
claude skill install https://github.com/codex-yolo/codex-yolo
# 方式二:手动配置 claude_desktop_config.json
{
"mcpServers": {
"codex-yolo": {
"command": "npx",
"args": ["-y", "codex-yolo"]
}
}
}
# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json
# 安装后在 Claude 对话中直接使用 # 示例: 用户: 请帮我用 Codex-YOLO 执行以下任务... Claude: [自动调用 Codex-YOLO MCP 工具处理请求] # 查看可用工具列表 # 在 Claude 中输入:"列出所有可用的 MCP 工具"
// claude_desktop_config.json 配置示例
{
"mcpServers": {
"codex-yolo": {
"command": "npx",
"args": ["-y", "codex-yolo"],
"env": {
// "API_KEY": "your-api-key-here"
}
}
}
}
// 保存后重启 Claude Desktop 生效
[!CAUTION] DO NOT USE THIS TOOL ON CORPORATE HARDWARE OR CONNECTED TO A CORPORATE NETWORK. This tool auto-approves all Codex CLI permission prompts without human review, including destructive commands. For maximum isolation, run it on a dedicated bare-metal server with no personal data, no saved credentials, and no access to sensitive networks. You accept full responsibility for any consequences.
git merge-tree across all branch pairs and logs conflicts as they emerge.--yolo, so this is intended only for isolated environments where broad command execution is acceptable.npm install -g @openai/codex)git merge-tree --write-tree)One-liner (macOS, Linux, WSL, Termux):
command -v curl >/dev/null || { s=; [ "$(id -u)" != 0 ] && s=sudo; command -v apt-get >/dev/null && { $s apt-get update && $s apt-get install -y curl; } || command -v dnf >/dev/null && $s dnf install -y curl || command -v yum >/dev/null && $s yum install -y curl || command -v apk >/dev/null && $s apk add curl || command -v pacman >/dev/null && $s pacman -S --noconfirm curl || command -v pkg >/dev/null && pkg install -y curl || command -v brew >/dev/null && brew install curl; }; curl -fsSL https://raw.githubusercontent.com/codex-yolo/codex-yolo/refs/heads/main/install.sh | bash && export PATH="${CODEX_YOLO_BIN_DIR:-$HOME/.local/bin}:${CODEX_YOLO_HOME:-$HOME/.codex-yolo}/bin:$PATH"
This clones to ~/.codex-yolo and symlinks the binary into ~/.local/bin. If ~/.local/bin is not writable, the installer falls back to ~/.codex-yolo/bin; you can also set CODEX_YOLO_BIN_DIR to choose a writable bin directory. It also installs git, tmux, curl, and codex if they are missing. Codex CLI is installed from the standalone GitHub release first, with npm as a fallback. Re-running the one-liner upgrades both codex-yolo and Codex CLI: if a newer stable Codex CLI release is published at openai/codex /releases/latest and the installed codex binary was placed by this script, it is overwritten with the new version (npm- or system-managed codex installs are left alone; set CODEX_YOLO_SKIP_CODEX_UPGRADE=1 to disable, or CODEX_YOLO_CODEX_VERSION=rust-v0.X.Y to pin a specific tag). Override the install location with CODEX_YOLO_HOME:
CODEX_YOLO_HOME="$HOME/my/path"; command -v curl >/dev/null || { s=; [ "$(id -u)" != 0 ] && s=sudo; command -v apt-get >/dev/null && { $s apt-get update && $s apt-get install -y curl; } || command -v dnf >/dev/null && $s dnf install -y curl || command -v yum >/dev/null && $s yum install -y curl || command -v apk >/dev/null && $s apk add curl || command -v pacman >/dev/null && $s pacman -S --noconfirm curl || command -v pkg >/dev/null && pkg install -y curl || command -v brew >/dev/null && brew install curl; }; curl -fsSL https://raw.githubusercontent.com/codex-yolo/codex-yolo/refs/heads/main/install.sh | CODEX_YOLO_HOME="$CODEX_YOLO_HOME" bash && export PATH="${CODEX_YOLO_BIN_DIR:-$HOME/.local/bin}:$CODEX_YOLO_HOME/bin:$PATH"
Local install (from a cloned repo; no network access needed if Codex CLI is already installed):
git clone https://github.com/codex-yolo/codex-yolo.git ~/.codex-yolo
cd ~/.codex-yolo
./install.sh --local
Manual install:
git clone https://github.com/codex-yolo/codex-yolo.git ~/.codex-yolo
ln -s ~/.codex-yolo/codex-yolo ~/.local/bin/codex-yolo
Then run from any project directory:
cd /path/to/your/project
codex-yolo "fix the tests" "update docs"
The tool runs agents in whatever directory you invoke it from (or the -d/--dir path if specified).
```bash
-s, --session NAME Custom tmux session name (default: codex-yolo-<timestamp>)
-d, --dir PATH Working directory for agents (default: current directory)
-m, --model MODEL Model to use (e.g., o4-mini, o3, gpt-4.1)
-p, --poll SECONDS Approver poll interval (default: 0.3)
-f, --file FILE Read a multiline prompt from a text file
-c, --command STRING Slash command to run in the control pane after launch
(e.g. "/loop 10m /plan ...", "/queue [...]", "/help").
Must start with '/'. Multi-line strings are sent as a paste.
Works with --resume to inject into an existing session.
-r, --resume Re-attach to an existing yolo session
--permissions PROFILE Set Codex /permissions profile (default: full-access when allowed, else auto-review)
--no-codex-sandbox Disable Codex sandboxing (for externally sandboxed containers)
--force-codex-sandbox Require Codex sandboxing; do not auto-fallback when unsupported
-h, --help Show help
Worktree options:
-w, --worktree Run each agent in its own git worktree
--base-branch BRANCH Base branch for worktrees (default: current branch)
--no-merge Skip auto-merge after agents complete
--no-cleanup Keep worktrees after merge
--conflict-poll SECS Conflict detection interval (default: 5)
install.sh options:
--local Install from the local repo without pulling from GitHub
By default, codex-yolo probes codex sandbox linux true once. In containers where bubblewrap fails with namespace permission errors, codex-yolo creates a temporary fake bwrap earlier in PATH and launches agents without Codex sandboxing. The fake bwrap executes the command after bubblewrap's -- separator directly, so it should only be used inside an externally isolated container. Use --force-codex-sandbox to require the real sandbox and surface failures instead.
For Codex /permissions, codex-yolo defaults to Full Access when the active Codex requirements allow it. If Full Access is disabled by requirements, it uses Auto-review (codex-auto-review). In containers where the Codex sandbox is unavailable and codex-yolo has to rely on external isolation, the auto default also uses Auto-review. Override this with `--permissions full-access, --permissions auto-review, or --permissions none`. For standard interactive Auto-review sessions, codex-yolo also reconciles the TUI once at startup so /permissions shows Auto-review (current).
When --worktree is enabled, three additional components run alongside the approver:
lib/worktree-manager.sh) creates a branch and git worktree per agent in <repo>-worktrees/<session>/.lib/conflict-daemon.sh) polls every --conflict-poll seconds and runs git merge-tree --write-tree across branch pairs. Conflicts are logged to the audit log.lib/merge-resolver.sh) waits for codex exec agents to finish, auto-commits uncommitted changes, merges branches into the base branch, and starts a Codex resolver task if a merge conflict occurs.自动化OpenAI Codex CLI代理工具,提高效率
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
AI Skill Hub 点评:Codex-YOLO 的核心功能完整,质量良好。对于Claude Desktop / Claude Code 用户来说,这是一个值得纳入个人工具库的选择。建议先在非生产环境试用,再逐步推广。
| 原始名称 | codex-yolo |
| 原始描述 | 开源MCP工具:Run parallel OpenAI Codex CLI agents in tmux with automatic permission approval.。⭐11 · Shell |
| Topics | ai-agentsauto-approveautomationbashcli-tool |
| GitHub | https://github.com/codex-yolo/codex-yolo |
| License | MIT |
| 语言 | Shell |
收录时间:2026-06-03 · 更新时间:2026-06-06 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端