Skip to content

Getting Started

Install

Grab the latest binary from GitHub Releases and put it on your $PATH:

tar xzf smelt-*.tar.gz
sudo mv smelt /usr/local/bin/

cargo install --git https://github.com/leonardcser/smelt.git

Run

Just run smelt. The first launch opens a provider wizard. Subscription providers log in with OAuth and can be used immediately; API-key providers write a starter ~/.config/smelt/init.lua.

You can also skip the wizard and configure everything from the command line:

smelt --model gpt-5.5 --api-base https://api.openai.com/v1 --api-key-env OPENAI_API_KEY
smelt --headless --model openai/gpt-5 "summarize @README.md"
smelt --resume              # open the saved-session picker

See the CLI Reference for every flag.

Subscription providers

These use your existing subscription and require no API key. Run smelt auth once, pick the provider, then start smelt with no extra flags:

smelt auth                # one-time browser or device-code login
smelt

smelt auth                # device-code login
smelt

After login, Smelt registers the kimi-code provider automatically. Start Smelt with no flags to use it.

smelt auth                # device-code login
smelt

Every model your Copilot account exposes (Claude, GPT, Grok, …) is available immediately.

API-key providers

These need --model and an environment variable or config file:

export OPENAI_API_KEY=...
smelt --model gpt-5.5 \
      --api-base https://api.openai.com/v1 \
      --api-key-env OPENAI_API_KEY

OpenRouter and other OpenAI-compatible services follow the same shape; set --api-base to the provider's base URL and swap the model name.

export ANTHROPIC_API_KEY=...
smelt --model claude-opus-4-8 \
      --api-base https://api.anthropic.com/v1 \
      --api-key-env ANTHROPIC_API_KEY

ollama pull qwen3.6:27b
smelt --model qwen3.6:27b --api-base http://localhost:11434/v1

Any OpenAI-compatible server works (Ollama, vLLM, SGLang, llama.cpp).

Save your config

Once you have a setup you like, write it to ~/.config/smelt/init.lua and run smelt from then on with no flags. Keeping config in a file means your providers, keymaps, and custom commands are version-controlled and portable across machines. You do not need to remember a long CLI invocation every time.

smelt.provider.register("ollama", {
  type = "openai-compatible",
  api_base = "http://localhost:11434/v1",
  models = { "qwen3.6:27b" },
})

smelt.provider.register("openai", {
  type = "openai",
  api_base = "https://api.openai.com/v1",
  api_key_env = "OPENAI_API_KEY",
  models = { "gpt-5.5" },
})

smelt.settings.vim = true

Switch models at runtime with /model. Edit the file and press F5 to hot-reload without losing the session.

init.lua is real Lua, not a schema: keymaps, slash commands, MCP servers, permission rules, statusline segments, and custom tools all live here.

Next