Skip to content
RunAPI Developer Docs
EN 中文
Developer Resources
Developer Resources

Jan

Add RunAPI as a Jan provider and verify a chat.

View Markdown

Connect Jan to RunAPI with a custom OpenAI-compatible provider. Create a standard RunAPI API key in the Authentication Guide before continuing.

Install Jan

Download and install the desktop application from the official Jan website, then open its settings.

Add the RunAPI provider

  1. Open Settings -> Model Providers -> Add Provider.
  2. Set the API format to OpenAI-compatible and the provider name to RunAPI.
  3. Set the Base URL to https://runapi.ai/v1, enter your RunAPI API key, and create the provider.
  4. Select a discovered model or add an exact identifier from the Model Catalog in the provider’s Models area.

The selected API format is OpenAI-compatible, so this Base URL includes /v1. A provider using another public protocol must use that protocol’s own client configuration instead.

When the provider is saved, Jan requests {base_url}/models, which resolves to https://runapi.ai/v1/models. If the model you need is not returned, use the manual + action instead of changing the Base URL.

Configure model capabilities

Jan does not infer every capability for a custom provider. Enable only the vision, tools, or audio behavior that the selected model declares in the Model Catalog. A UI switch cannot add an unsupported API capability.

Verify the connection

Start a new chat, select the RunAPI model, and send Reply with the word connected. A normal complete or streamed response confirms the provider and model are active.

Troubleshoot

  • Every request returns 404: make sure the Base URL includes /v1; Jan appends the operation path.
  • No models appear: confirm the Base URL is https://runapi.ai/v1, then add the exact model identifier manually.
  • Provider creation returns 401 or 403: replace the key with a valid standard RunAPI API key, not a Management Key.
  • Capability controls do not match the model: configure each custom model from its declared capabilities rather than relying on automatic detection.
  • A fallback key is not tried: Jan rotates keys for 401, 403, or 429; other errors do not trigger key rotation.

For shared protocol behavior, use the LLM API Quickstart. For request and response fields, use the Chat Completions API Reference.