LlamaStash

Zero-overhead, terminal-native local-LLM launcher.

A fast TUI and CLI with init wizard for launching local LLMs. One Rust binary that's a TUI, a CLI, a daemon, and an OpenAI-compatible proxy. llama.cpp is the direct, zero-overhead backend (vs raw llama-server), behind a pluggable backend seam so other engines can plug in.

LlamaStash gives you one binary for three jobs:

• a keyboard-driven TUI
• a scriptable CLI with stable --json output
• an on-demand daemon that supervises running models

Install

# macOS + Linux
curl -fsSL https://llamastash.dev/install.sh | sh

# Homebrew
brew install llamastash/llamastash/llamastash

# Cargo
cargo install llamastash

Then run:

llamastash init

The init wizard detects your hardware, installs the right llama-server, downloads a starter GGUF, writes a tuned config, and smoke-launches it.

What you get

• hardware-aware first-run setup
• auto-discovery of HuggingFace, Ollama, and LM Studio model caches
• a fast TUI for launching, testing, and stopping models
• an OpenAI-compatible local proxy for external tools and agents
• CLI + JSON contracts for scripts and coding agents

Repositories

• llamastash/llamastash - main source repo
• llamastash/homebrew-llamastash - Homebrew tap
• llamastash/llamastash.github.io - website and install-script mirror

Links

• Website
• Install guide
• Usage docs
• Issues