Configure providers

The guided onboarding flow is the recommended way to configure a provider:

odinn onboard

Use the explicit commands below for headless systems, scripts, and advanced setups.

Inspect the catalog

odinn config provider catalog
odinn config provider list

The catalog contains built-in presets. A preset supplies an authentication mode, endpoint shape, and model defaults; it does not guarantee that a remote provider or model is currently available.

API-key provider

Set the provider’s documented environment variable, then onboard:

export OPENAI_API_KEY="..."
odinn onboard --provider openai --auth api-key

Do not place API keys in shell scripts, screenshots, plans, audit payloads, or the ordinary configuration file.

OAuth provider

For OpenAI/ChatGPT browser sign-in, choose OpenAI in guided onboarding. The flow opens a browser and stores refreshable OAuth material in the owner-only OAuth store.

Imported OpenClaw OAuth can be copied without changing the source installation:

odinn auth import openclaw

Local provider

Start the local server first, then use the exact model name it exposes:

odinn onboard --provider ollama --model <installed-model>

Ódinn also provides presets for LM Studio, vLLM, SGLang, and LiteLLM.

Add a custom compatible endpoint

odinn config provider add <name> \
  --base-url <https-or-loopback-url> \
  --model <model> \
  --api-key-env <ENVIRONMENT_VARIABLE>

Use HTTPS for remote endpoints. Private-network access remains subject to the configured policy.

Select and verify

odinn config model list
odinn config model default <provider:model>
odinn onboard --verify --non-interactive

A verification failure should be treated by category: credentials, quota, model availability, timeout, local-server availability, or network failure. See Troubleshooting.