Model providers

Rusty does not include a model. You connect the providers you already use in the LLM Connection Center, opened from LLM Integrations in the left rail. Requests go directly to the providers you configure.

Built-in providers

Provider	How you connect	Notes
OpenAI	API key	Connects directly to OpenAI. If the key is blank, Rusty falls back to the OPENAI_API_KEY environment variable.
Anthropic	API key	Connects directly to Anthropic's Claude API. Falls back to ANTHROPIC_API_KEY if the key is blank.
OpenRouter	API key	A unified OpenAI-compatible gateway to many providers. Falls back to OPENROUTER_API_KEY. Use Fetch Models to discover what is available. Required for automatic model selection.
GitHub Models	Personal access token	Uses GitHub's OpenAI-compatible inference API. Needs a token with the models:read scope, or GITHUB_TOKEN.
GitHub Copilot	Sign-in	Uses your Copilot subscription through GitHub's official Copilot SDK. Available models and premium-request usage depend on your plan and organization policy.
OpenAI Codex	Sign-in	Uses your existing OpenAI Codex sign-in through the official Codex app-server. Credentials are shared with the Codex CLI and desktop app through ~/.codex. Available models come from your account.
OpenCode Zen	API key	Connects to OpenCode Zen and uses OPENCODE_API_KEY if the key is blank. To use OpenCode Go, register it separately as opencode-go with its own base URL.

Important

Claude models are reached with an Anthropic API key. Signing in with a Claude subscription is not supported.

Connect a provider with an API key

1. Open LLM Integrations.

2. Select the provider in the list. Its Provider Configuration panel opens.

3. Fill in the fields:

   Field	What it is
   API Authorization Key	Your key or token. It is stored encrypted, see Data, storage & privacy.
   Connection Base URL	The API endpoint, for example https://api.openai.com/v1. Pre-filled for built-in providers.
   Model Catalog URL	Where Rusty lists models. Defaults to the base URL plus /models.
   Default API Protocol	How Rusty talks to the API. See below.
   Default Target Model	The model used when none is chosen explicitly.

4. Click Save Configuration.

5. Fetch the model catalog so the provider's models show up in the model pickers.

Sign in with GitHub Copilot or OpenAI Codex

These two providers use your existing subscription instead of an API key.

1. Select GitHub Copilot or OpenAI Codex in the list.
2. Click Sign in with GitHub Copilot (or Sign in with OpenAI Codex).
3. Complete the sign-in. Rusty either opens a browser window for you, or shows a device code: use Copy Code, then open the verification page and enter it.
4. When it reads Connected, the models your account can use appear in the pickers.

Use Cancel to abandon a sign-in in progress, Re-authenticate to sign in again, and Sign out to disconnect.

Rusty never stores the OAuth token in its own provider settings. The official Codex app-server keeps Codex credentials in ~/.codex, and the Copilot CLI uses the system credential store.

The first time you sign in, Rusty downloads that provider's runtime, since installers do not include it. See Installation. If something goes wrong, expand Authentication diagnostics and use Copy Diagnostics to copy details you can share.

Add a custom or local provider

Any server that speaks an OpenAI-compatible API works, including Ollama, LM Studio, vLLM and LiteLLM gateways.

1. In the LLM Connection Center, choose Register Custom LLM / Local Host (New Custom Provider).

2. Fill in the form:

   Field	Example
   Provider ID	ollama
   Display Name	Local Ollama Runner
   Base API URL	http://localhost:11434/v1
   Catalog URL (optional)	Defaults to {base URL}/models
   API Protocol	See below
   Authentication	See below
   Models (comma-separated)	qwen2.5-coder:7b, llama3.3:70b

3. Click Register Provider.

API protocols

Option	Use it for
OpenAI Chat Completions	Most OpenAI-compatible servers: Ollama, LM Studio, vLLM, LiteLLM, OpenRouter.
OpenAI Responses	Endpoints that implement OpenAI's Responses API.
Anthropic Messages	Anthropic's API or a compatible gateway.
Google Generative AI	Google's Gemini API.

Authentication

Option	Meaning
None / Local	No credentials, for example a local server.
Bearer token	Sends your key as Authorization: Bearer ….
Anthropic x-api-key	Sends your key in Anthropic's x-api-key header.

Reasoning effort variants

For models that support it, Rusty lists reasoning-effort variants of the same model: Minimal, Low, Medium, High and Extra High. Pick a variant in the model picker like any other model, and use them when you map models to AUTO levels.

Environment variables

Where a provider says it falls back to an environment variable, that variable is read from the environment Rusty was launched with. Apps started from Finder or the Dock do not inherit variables you export in your shell profile, so paste the key into the provider's configuration instead.