Configuration

Providers & OAuth

minnal speaks directly to model providers — including the ones you already pay a subscription for. No proxy, no vendor in the middle.

Use minnal login --provider <name> to start an OAuth or device-code flow that stores credentials in your OS keyring.

Command	Provider
`minnal login --provider claude`	Anthropic Claude (Pro / Team / Console)
`minnal login --provider openai`	OpenAI / ChatGPT / Codex
`minnal login --provider gemini`	Google Gemini
`minnal login --provider copilot`	GitHub Copilot
`minnal login --provider azure`	Azure OpenAI
`minnal login --provider alibaba-coding-plan`	Alibaba Cloud Coding Plan
`minnal login --provider fireworks`	Fireworks AI
`minnal login --provider minimax`	MiniMax
`minnal login --provider lmstudio`	LM Studio (local)
`minnal login --provider ollama`	Ollama (local)
`minnal login --provider openai-compatible`	Custom OpenAI-compatible endpoint

Self-hosted & OpenAI-compatible endpoints

Add an entry under [providers.openai_compatible.<name>] in ~/.minnal/config.toml:

toml

[providers.openai_compatible.local-vllm]
base_url = "http://localhost:8000/v1"
model    = "qwen2.5-coder-32b"
api_key  = ""                       # optional

[providers.openai_compatible.together]
base_url = "https://api.together.xyz/v1"
model    = "deepseek-ai/DeepSeek-V3"
api_key  = "${TOGETHER_API_KEY}"

Switching models inside a session

/models                # picker
/model gpt-5.5
/model claude-opus-4.7

Cache-warmth notes

Anthropic prompt caching expires after 5 minutes of inactivity. minnal warns you in the UI when the cache has gone cold and notifies you on unexpected cache misses, so you do not silently overpay.

Currently supported providers