OpenProviderOpenProvider documentation
Set up providers, browse free models, use the app, and integrate with the OpenAI-style API.
OpenProvider is a free-model gateway. You connect provider credentials once, OpenProvider syncs a free-only model catalog, and your apps call one OpenProvider surface for chat, image generation, image analysis, and speech.
The web app is for daily work: explore models, save useful routes, chat with text models, test image and speech routes, and manage provider/API credentials in Account.
Browse synced free models across text, image generation, image analysis, and speech categories.
Save provider credentials once. OpenProvider keeps those provider-native keys on the server.
Use auto routing for fallback, or pass an exact model id when you need deterministic provider behavior.
Client apps call OpenProvider with generated keys instead of handling provider credentials directly.
From empty workspace to working route
A working OpenProvider workspace needs a signed-in account, at least one configured provider, a synced catalog, and an OpenProvider API key for external apps.
- 01Sign in
Use the account menu to create a personal workspace for provider keys, API keys, saved models, and chat history.
Open account - 02Connect providers
Open Account -> Provider setup and save keys for the free providers you want OpenProvider to route through.
Open provider setup - 03Check the catalog
Open Models to confirm synced free models, filter by category, and inspect exact model ids.
Open models - 04Generate an API key
Open Account -> API keys and create an OpenProvider key for your application.
Open API keys
What each account area is for
Account is the operational center for credentials, API keys, saved model shortcuts, and conversation history.
Add, update, and keep encrypted provider credentials tied to your workspace.
API keysCreate and revoke OpenProvider keys for apps that call `/v1/*` routes.
Saved modelsBookmark models from the catalog and keep shortcuts for future routes.
ConversationsReview recent chat history saved in your personal workspace.
Understand the model catalog
Models shows the current free registry after provider discovery and free-only filtering. Use it to compare providers, inspect context/output limits, save models, and copy exact model ids for API calls.
Text, image, and speech are first-class catalog categories. Image analysis models are routed through the analysis workflow.
The Working filter uses provider status checks so you can avoid routes that recently failed.
Saved models become workspace shortcuts for repeated testing and app integration.
Use the chat console
Chat is the fastest way to test text models. Start with OpenProvider Auto Free for fallback across configured providers, or switch to a specific model when you need predictable provider behavior.
Chat -> select model -> send prompt -> save conversation in AccountGenerate images, analyze images, and create speech
The playground is for media workflows. Image routes generate pictures from prompts, image analysis accepts URLs or uploads, and speech routes return audio bytes from text prompts.
Choose auto routing or exact model ids
Auto routing is best when availability matters. Exact model ids are best when your app depends on a specific provider, context window, category, or output behavior.
| Mode | Use when |
|---|---|
| Auto route | You want OpenProvider to select an available free route and fall back when one provider fails. |
| Exact model | You need deterministic provider behavior, model-specific limits, or a known output style. |
| Provider setup | You want to expand route availability by connecting more provider keys. |
Use OpenProvider API keys
Generate an OpenProvider key from Account, then send it as a Bearer token. Provider-native credentials never need to leave the OpenProvider server.
Authorization: Bearer opk_live_...OpenAI-style routes for supported workflows
All `/v1/*` routes require an OpenProvider API key. Provider credentials stay encrypted in the workspace.
/v1/modelsReturns the free model registry for the authenticated workspace.
/v1/providers/statusReturns configured provider state, discovery errors, and catalog coverage.
/v1/chat/completionsOpenAI-compatible chat completions with OpenProvider auto fallback.
/v1/images/generationsText-to-image generation through configured free image providers.
/v1/images/analyzeImage understanding from an image URL, data URL, or multipart upload.
/v1/audio/speechText-to-speech that returns raw audio bytes.
Interactive request builder
Choose a workflow below to generate copyable cURL and JavaScript snippets, edit the request body, and run a live request with your OpenProvider key.
Chat
OpenAI-compatible chat completions with OpenProvider auto fallback.
/v1/chat/completionscurl -X POST "https://your-openprovider.app/v1/chat/completions" \
-H "Authorization: Bearer opk_live_..." \
-H "Content-Type: application/json" \
--data '{
"model": "openprovider/auto-free",
"messages": [
{
"role": "user",
"content": "Explain OpenProvider auto routing in one paragraph."
}
],
"temperature": 0.3
}'Run an example to inspect the response here.One error shape across providers
OpenProvider normalizes route, model, provider, and authentication failures into a predictable JSON envelope.
Missing or invalid OpenProvider API key.
Exact model is not in the free registry or does not match the endpoint category.
No configured free model is currently available for the requested workflow.
A provider free tier returned a rate-limit response. Retry later or route to another model.