setup-ai
Cross-OS installer for an AI coding toolchain. One command bootstraps pi, Claude Code, Codex, Antigravity, , , , the dotfiles, , and the skill — each via its method for your OS.
接入你的智能体
发布者自己的命令。不是 PlugMCP 端点。
用我们的密钥与日志托管一份副本。
npx -y @darkrei08/[email protected]Claude Code 没有针对本地命令的安装链接。复制下方命令并在终端中运行。
claude mcp add darkrei08-setup-ai -- npx -y @darkrei08/setup-ai@4.0.3在终端中运行。
{ "mcpServers": { "darkrei08-setup-ai": { "command": "npx", "args": [ "-y", "@darkrei08/[email protected]" ] } }}claude_desktop_config.json
远程 URL 还是本机命令? 两种安装方式的区别
来自发布者
Cross-OS installer for an AI coding toolchain. One command bootstraps pi, Claude Code, Codex, Antigravity, , , , the dotfiles, , and the skill — each via its method for your OS.
Cross-OS installer for an AI coding toolchain (pi, codex, antigravity, opencode, herdr, gentle-ai/gga, Engineering Excellence) with an interactive module menu and per-agent MCP registration.
One command — opens a menu to choose what to set up:
npx @darkrei08/setup-ai
Non-interactive:
npx @darkrei08/setup-ai --all --verbose --yes # recommended: everything, unattended
npx @darkrei08/setup-ai --only pi,codex,opencode # just these
npx @darkrei08/setup-ai --yes # core set, no prompts
npx @darkrei08/setup-ai --all # everything
npx @darkrei08/setup-ai --verbose # core set, no prompts, with live command logs
npx @darkrei08/setup-ai --dry-run # show the plan without changing anything
npx @darkrei08/setup-ai --uninstall [--yes] [--only ...] [--purge]
Startup acceptance check: bash tests/pi-startup-check.sh.
JSONL schema check: bash tests/jsonl-schema.sh.
Three details worth knowing:
npx github:darkrei08/setup-ai also works, but many npm setups block git fetches (EALLOWGIT); the published package avoids that. Run it from any folder except the repo's own source dir.\). If node or git was just installed, open a new terminal so PATH refreshes.setup-ai.sh (macOS/Linux) or setup-ai.ps1 (Windows). Each module and command is shown with live output, so prompts such as sudo remain visible; --list shows modules; menu: pick a preset (Core, Everything, Custom, Minimal, Gentle AI), drill into categories with Enter, then confirm the summary; keys: arrows/j/k move · Space toggle · a all/none · Esc back · q quit. Gentle AI installs only the base, Node, Pi, pi-extensible-workflows, and gentle-ai modules.--dry-run writes nothing, never prompts, and discards its log/report when the plan ends. --uninstall prints the inventory first and requires --yes; destructive items additionally require --purge. The uninstall path never touches ~/.pi/agent/auth.json or ~/.pi/agent/sessions/.For a one-time installer, npx is ideal — nothing is left behind.
logs/setup_<runid>.log, one line per event
(<timestamp> [LEVEL] <phase> <event>: <message>), alongside the structured
logs/setup_<runid>.jsonl (one JSON record per event) and the
logs/engineering-report_<runid>.md. --verbose renders the same events live.
command_start → command_success → step_result is one successful step,
immediately followed by the next step's command_start: normal, not an error.
Reference: docs/logs.md..jsonl ends with one run_summary record (outcome,
per-module status, step counts); a one-liner to read it back is in
docs/logs.md.~/.pi/agent/settings.json,
skills as <skill-root>/<skill>/SKILL.md for every targeted agent, installed
npm trees as <dir>/node_modules/<pkg>/package.json.~/.pi/agent/settings.json (there is no
~/.pi/agent.json). The dotenv module's setup_env.sh populates ~/.pi/agent
with a selective rsync from the dotenv checkout's pi/agent directory
(settings, models, prompts, skills, themes, ...), leaving auth.json,
sessions/ and pi-managed installs untouched. It is a real directory, not a
symlink into the checkout.Everything below the quick start is split into focused pages. Start from the one that matches your question.
pi-packages.example.txt in the package root is the starting point for the
declarative Pi package manifest.
MIT (this installer). Installed tools keep their own licenses — note cockpit-tools is CC BY-NC-SA 4.0 (non-commercial).
Cross-OS installer for an AI coding toolchain (pi, codex, antigravity, opencode, herdr, gentle-ai/gga, Engineering Excellence) with an interactive module menu and per-agent MCP registration.
{ "mcpServers": { "darkrei08-setup-ai": { "command": "npx", "args": [ "-y", "@darkrei08/[email protected]" ] } }}工具
暂无工具列表。远程服务器在首次健康检查后会获得工具列表。
常见问题
在本页复制适用于 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 配置。它使用发布者的安装方式,而不是 PlugMCP 代理。
此记录未列出必需的环境变量。
@darkrei08/setup Ai 通过 stdio 在本地运行。
上次检查时间:2026-10-28。 Cross-OS installer for an AI coding toolchain (pi, codex, antigravity, opencode, herdr, gentle-ai/gga, Engineering Excellence) with an interactive module menu and per-agent MCP registration.