Provider setup

Open the terminal UI and run the setup wizard:

supercharge

Then type:

/setup

/login opens the same wizard. It does not sign you into a hosted Supercharge or Grok account. /logout clears session-based auth only. The provider URL and key stay under /setup.

What you enter

1. Pick a listed provider, or choose Custom and type a base URL.
2. Choose the API format: Chat Completions, Messages, or Responses.
3. Enter the API key when the provider requires one. A local server can run without a key.
4. Pick a model from the list the provider returns, or type an id that endpoint accepts.

The base URL should look like https://api.example.com/v1. Do not append /chat/completions, /messages, or /responses. The CLI adds that path from the format you chose.

Saved config goes in ~/.supercharge/config.toml and is reused next launch. A global environment variable is not accepted as the provider API key. Put the key in /setup.

Formats

Change the saved format later without re-entering the key:

/format

If no valid setup exists yet, /format tells you to run /setup first. The endpoint must actually support the format you pick.

Switch models later

/model
/model your-model-id
/m your-model-id
/effort high

/model with no argument fetches /models from the saved base URL and lists chat models. The second argument on a reasoning model is an effort level: low, medium, high, or xhigh. /effort changes effort on the current model only when that model advertises it.

From the shell, for one run:

supercharge -m your-model-id
supercharge -p "Summarize this repo" -m your-model-id
supercharge models

Ctrl+M in the scrollback opens the model picker. With the prompt focused, Ctrl+M toggles multiline input instead.

A persistent default:

[models]
default = "your-model-id"

Request size

max_request_bytes caps one request to your provider. Set it on a model, or on the provider so models inherit it. When it is unset, chat and responses requests cap at 50 MiB and the Messages API caps at 30 MiB. The TOML key is max_request_bytes.

Extra headers and query params

[model.my-model]
model = "model-id"
base_url = "https://api.example.com/v1"
name = "Display name"
api_backend = "chat_completions"
api_key = "enter-this-through-setup"
temperature = 0.7
top_p = 0.95
max_completion_tokens = 8192
context_window = 128000
extra_headers = { "x-api-key" = "sk-..." }
query_params = { api-version = "2026-07-22" }

extra_headers are sent as written on every request to that endpoint. Per-model values win over the same keys under [models].

Image and video providers

Chat setup does not configure images. Use a separate wizard:

/isetup
/imodels flux
/imagine a diagram of the request flow
/imagine-video a short pan across a desk

Aliases: /imagine-setup, /image-setup, /imagine-model, /image-model, /imodel. Image settings are [image_provider] in config.toml. /imagine needs /isetup first.