# Models Seris uses the model you bring. Open Settings → Models, choose a provider, enter its API key (local servers don't need one) and pick a default model. Requests go from your Mac straight to that provider. ## Providers | Provider | What you enter | | --- | --- | | Anthropic | API key | | OpenAI | API key | | Google Gemini | API key | | DeepSeek | API key | | Moonshot AI / Kimi (global, China and Kimi Coding) | API key | | OpenRouter | API key | | OpenAI compatible | Endpoint URL, API key and model | | Anthropic compatible | Endpoint URL, API key and model | | Ollama | Nothing: local server at `http://127.0.0.1:11434/v1` | | LM Studio | Nothing: local server at `http://127.0.0.1:1234/v1` | ## Local models Start Ollama or LM Studio before you open Settings → Models. Seris finds the models the running server offers, so you can choose one without an API key, and requests stay on your machine. ## Compatible endpoints Choose OpenAI compatible or Anthropic compatible to use another service or a self-hosted gateway. Enter its endpoint URL, API key and model. OpenAI-compatible endpoints can use the Chat Completions or the Responses API. ## Keys and costs Desktop API keys are saved in the operating system credential store, not in a Seris account. Model requests use the provider's pricing and data policies, and testing a connection sends one short request. ## Environment variables for the standalone gateway When you run the gateway from source, you can configure credentials in the UI or with provider environment variables: ``` ANTHROPIC_API_KEY or ANTHROPIC_AUTH_TOKEN OPENAI_API_KEY GEMINI_API_KEY or GOOGLE_API_KEY OPENROUTER_API_KEY DEEPSEEK_API_KEY MOONSHOT_API_KEY, MOONSHOT_CN_API_KEY KIMI_API_KEY ``` Docs: https://seris.im/docs/