One-shot local LLM bootstrapper. Sets up everything you need to run a local LLM — automatically.

curl -fsSL https://raw.githubusercontent.com/shi3z/mubo/main/setup.sh | bash

Or clone and run:

git clone https://github.com/shi3z/mubo.git
cd mubo
./setup.sh

After setup, open http://localhost:8392 in your browser to use the AI agent.

• Fully automatic setup — Detects OS, CPU, RAM, GPU and configures everything
• Smart hardware detection — NVIDIA / AMD / Apple Silicon GPU, unified memory, NPU
• Optimal context length — Automatically set based on available memory (16K–128K)
• Safe optional installs — vLLM, MLX, Docker, uv are installed only when possible, never breaking the core
• Self-modifying AI agent — Web-based agent that can rewrite its own source code, with full version history and rollback

1. Environment detection — OS, CPU architecture, RAM, GPU type
2. Ollama install — Install Ollama and verify server health
3. Base model download — Pull gpt-oss:20b
4. Extended model creation — Create a derived model with optimal context length
5. Extras (optional) — vLLM, MLX, uv, Docker
6. Mubo Agent deploy — Launch the self-modifying AI agent with web UI

• bash
• Internet connection
• Sufficient storage for model weights

MIT