Install

Pick whichever path is most natural for you. All three give you the same experience.

Prerequisites

Before installing mini-claude, you need an inference server. The simplest is Ollama:

# macOS
brew install ollama
ollama serve &

# Linux
curl -fsSL https://ollama.com/install.sh | sh

# Pull a small model
ollama pull llama3.2:3b

Any OpenAI-compatible server works (llama.cpp, LM Studio, vLLM, …). You just need an HTTP endpoint exposing /v1/chat/completions and /v1/models.

From source (recommended)

You need Go 1.23 or later.

git clone https://github.com/hugostarte/Mini-Claude.git
cd Mini-Claude
go run ./cmd/tui

To build a standalone binary:

go build -o mini-claude ./cmd/tui
./mini-claude

Docker Compose

The repo ships a docker-compose.yml that orchestrates Ollama and the TUI. Because TUIs need a real terminal:

docker compose run --rm tui

Not docker compose up — the TUI needs a TTY which up doesn’t provide. The compose file sets tty: true and stdin_open: true on the tui service.

Prebuilt binary

Once we cut the first release, prebuilt binaries for macOS, Linux, and Windows will be attached to the GitHub release page.

# macOS arm64 example (coming soon)
curl -L https://github.com/hugostarte/Mini-Claude/releases/latest/download/mini-claude-darwin-arm64 -o mini-claude
chmod +x mini-claude
./mini-claude

Verifying it works

When mini-claude launches, you should see the welcome screen with the ASCII logo, your model, server URL, and the available commands. If you see “upstream 404” at the bottom, Ollama probably isn’t running or the model name doesn’t match — see Troubleshooting.

Configure