Providers & models

The full provider matrix, how to select a model, the credential chain, and pointing Stella at a local OpenAI-compatible server.

Stella is model-agnostic. It speaks each provider's native wire dialect and auto-detects which provider to use from the API keys you have set.

This page is the quick reference. For per-provider deep dives — setup walkthroughs, provider-specific configuration, and slugs — see API Providers, and for choosing between models (pricing, release dates, what each is best at) see the model guide.

Supported providers

Provider	id	Primary env var	Default model	Base URL	Dialect
Anthropic (Claude)	`anthropic`	`ANTHROPIC_API_KEY`	`claude-fable-5`	`https://api.anthropic.com`	Anthropic Messages
OpenAI (GPT)	`openai`	`OPENAI_API_KEY`	`gpt-5.5`	`https://api.openai.com/v1`	OpenAI Responses
Google Gemini	`gemini`	`GEMINI_API_KEY` (alias `GOOGLE_API_KEY`)	`gemini-3-pro`	`https://generativelanguage.googleapis.com/v1beta`	Gemini generateContent
Google Vertex AI	`vertex`	`VERTEX_ACCESS_TOKEN`	`gemini-3-pro`	project/region-scoped	Gemini generateContent
Amazon Bedrock	`bedrock`	`AWS_ACCESS_KEY_ID`	Claude via Converse	region-scoped	Bedrock Converse
xAI (Grok)	`xai`	`XAI_API_KEY`	`grok-4`	`https://api.x.ai/v1`	OpenAI-compatible
DeepSeek	`deepseek`	`DEEPSEEK_API_KEY`	`deepseek-chat`	`https://api.deepseek.com/v1`	OpenAI-compatible
Z.ai (GLM)	`zai`	`ZAI_API_KEY`	`glm-5.2`	`https://api.z.ai/api/paas/v4`	OpenAI-compatible
OpenRouter	`openrouter`	`OPENROUTER_API_KEY`	`openrouter/auto`	`https://openrouter.ai/api/v1`	OpenAI-compatible
Local	`local`	none (optional `LOCAL_API_KEY`)	you choose	pass `--base-url`	OpenAI-compatible

List them at any time — with live key status and effective base URLs:

stella models

Selecting a model

Pin the worker model for a single invocation with --model provider/model_id, or set the STELLA_MODEL environment variable for the whole shell:

stella --model zai/glm-5.2 run "fix the failing test"
export STELLA_MODEL=anthropic/claude-fable-5

With no --model, Stella auto-detects: it selects the first provider that has a resolvable credential, in preference order — zai, anthropic, openai, xai, deepseek, gemini, openrouter, then vertex and bedrock last.

Vertex and Bedrock sit last in auto-detection on purpose: they key off generic cloud credentials (VERTEX_ACCESS_TOKEN, AWS_ACCESS_KEY_ID) that are often present for unrelated reasons. Auto-detection never prefers them over an explicitly-configured provider — pin them with --model vertex/… or --model bedrock/… when you want them.

The credential chain

For the selected provider, Stella resolves the API key in this order — first hit wins:

--api-key flag (needs an explicit --model provider/...)
The provider's env var (and aliases, e.g. GOOGLE_API_KEY for Gemini)
settings.json providers.<id>.api_key
~/.config/stella/credentials.toml
Interactive prompt (on a TTY) — the entered key is saved to credentials.toml, so you are only prompted once

See Credentials for the file format and precedence details.

Prefer an env var, settings.json, or credentials.toml over --api-key for anything long-lived — a flag value is visible in shell history and in ps output.

Cloud providers with extra configuration

Vertex AI also needs a project — set VERTEX_PROJECT_ID (or GOOGLE_CLOUD_PROJECT), and optionally VERTEX_LOCATION (defaults to global). The access token is typically export VERTEX_ACCESS_TOKEN=$(gcloud auth print-access-token).

Amazon Bedrock needs AWS_SECRET_ACCESS_KEY alongside AWS_ACCESS_KEY_ID, plus an optional AWS_SESSION_TOKEN and AWS_REGION / AWS_DEFAULT_REGION (defaults to us-east-1).

Local / any OpenAI-compatible server

The local pseudo-provider points at any OpenAI-compatible endpoint — Ollama, vLLM, LM Studio, or a llama.cpp server. It is never auto-detected and needs no API key:

stella --model local/llama3.3 --base-url http://localhost:11434/v1 chat

The id local is reserved and cannot be redefined in settings.json — --base-url (or its env form STELLA_BASE_URL) is always required with --model local/<model>. To avoid retyping the endpoint, either export STELLA_BASE_URL, or define your own custom provider under a non-reserved id in settings.json — e.g. providers.myollama with a base_url (its dialect defaults to openai-compatible) — and select it with --model myollama/<model>.

Overriding a provider's defaults

Any built-in provider's base URL, key, display name, or default model can be overridden in settings.json — the recommended way to, for example, route Z.ai through its coding-plan endpoint without a provider-specific environment variable. For personal routing like a subscription endpoint, put the override in user scope (~/.config/stella/settings.json): project-scope base_url/api_key overrides are held back by the trust boundary until you set STELLA_TRUST_PROJECT=1. Ready-to-paste setups, organized by use case, live in Examples & recipes — including the Z.ai coding-plan recipe.