LLM Launchpad — AI skill for Claude Code
一条命令在 Apple Silicon Mac 上部署本地大模型(Qwen3.8-27B · Ollama · Claude Code 兼容).
How to install LLM Launchpad
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open majiayu000/llm-launchpad and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What LLM Launchpad does
一条命令在 Apple Silicon Mac 上部署本地大模型(Qwen3.8-27B · Ollama · Claude Code 兼容).
Alternatives in AI
- Qwen Code — A command-line AI workflow tool adapted from Gemini CLI, optimized for Qwen3-Coder models with enhanced parser 20.8k ★
- Opencodex — Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama… 12.3k ★
- CodeIsland — Real-time AI coding agent status panel in your MacBook notch — live status, approvals & replies for 13 AI tool 2.3k ★
README
Qwen3.8-27B · 本地一键部署(Ollama)
在你的 Apple Silicon Mac 上一条命令跑起 Qwen3.8-27B,提供 OpenAI 兼容 API,并可直接作为 Claude Code 或 Codex CLI 的后端模型。全程只监听 `127.0.0.1`,数据不出本机。
硬件要求
| 项目 | 要求 |
|---|---|
| 芯片 | Apple Silicon(M1/M2/M3/M4 系列) |
| 统一内存 | 24GB 起步(16–24GB 安装时会要求确认;低于 16GB 拒绝安装) |
| 磁盘空闲 | 40GB(模型约 17GB + 运行余量) |
| 软件 | macOS + Homebrew(ollama 未安装时会自动 brew install) |
默认使用 MLX 引擎的 4bit 量化版(`qwen3.8:27b-mlx`,约 18GB),加载后约占 20GB 内存。参考速度:M2 Max 96GB 上生成约 16 token/s;改用 GGUF 版(`QWEN38_MODEL=qwen3.8:27b`)约 8 token/s,但新开对话的首次响应更快。
一键安装
git clone https://github.com/majiayu000/llm-launchpad.git
cd llm-launchpad
./install.sh
`install.sh` 幂等可重跑:预检硬件 → 准备 ollama → 启动 launchd 常驻服务 → 下载模型(断点续传)→ 发一条真实请求冒烟测试 → 打印使用入口。
先不确定机器能不能跑?用干跑模式只做预检:
./install.sh --check
四种使用方式
**1. 命令行对话**(thinking 默认开启,`reasoning_effort=medium`)
./scripts/chat.sh '解释一下什么是向量数据库'
**2. Claude Code 直接用本地模型**
./scripts/claude-code.sh
**3. Codex CLI 直接用本地模型**
./scripts/codex.sh
入口使用独立的 `~/.local/share/qwen38-ollama/codex`,不会改动现有 `~/.codex`。默认保留命令确认和工作区沙箱。
已安装其他 Ollama 模型时可显式选择,包括用户自行评估的 uncensored 模型;这只改变模型,不会关闭 Codex 的命令审批或沙箱:
QWEN38_MODEL=example/uncensored:27b ./scripts/codex.sh
**4. OpenAI SDK / 任意兼容客户端**
from openai import OpenAI
client = OpenAI(base_url="http://127.0.0.1:11439/v1", api_key="local")
response = client.chat.completions.create(
model="qwen3.8:27b-mlx",
messages=[{"role": "user", "content": "你好"}],
)
print(response.choices[0].message.content)
Anthropic Messages 兼容入口在 `http://127.0.0.1:11440`。请求必须带与 `QWEN38_COMPAT_TOKEN`(默认 `ollama`)一致的 `x-api-key` 或 `Authorization: Bearer`;缺失或不匹配返回 401。`scripts/claude-code.sh` 会读取同一环境变量。
curl http://127.0.0.1:11440/v1/messages \
-H 'Content-Type: application/json' \
-H 'x-api-key: ollama' \
-H 'anthropic-version: 2023-06-01' \
-d '{
"model": "qwen3.8:27b-mlx",
"max_tokens": 64,
"messages": [{"role": "user", "content":
Related Skills
Ovo Local LLM
A private Claude-Code-style coding agent for Apple Silicon — run chat, code, and local model workflows on-devi
Flow LLM
Local LLM gateway for Apple Silicon. Works with OpenClaw, Hermes Agent, Claude Code, and Codex (AIRun). No Oll
Mac Code
mac code — Claude Code, but it runs on your Mac for free. 35B AI agent at 30 tok/s via Apple Silicon flash-pag
Qwopus
Local AI coding agent powered by Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled. Like Claude Code, but runs e
Vllm Mlx
High-performance OpenAI and Anthropic compatible LLM inference server for Apple Silicon. Native MLX, continuou
Wax
Shared Single-file memory layer for all your agents, sub mili-second RAG over text, photo and video on Apple S
Related Agents
Buildhost Lead
Use PROACTIVELY and automatically — do not wait to be asked — as the Claude-Code node lead FOR an Apple-silico
ZeroClaw Android
Run AI agents 24/7 on your Android phone. Native Rust core, 25+ providers (OpenAI, Claude, Gemini, Groq, DeepS
Delegate
Expert LLM delegation specialist that seamlessly connects to external language models including GPT-4, GPT-3.5