Any provider supporting the OpenAI chat completions API or Anthropic Messages API can be used. Enter a custom endpoint in the provider settings.
Cloud Providers
| Provider | Protocol | Base URL | Popular Models | Status |
|---|---|---|---|---|
| OpenAI | OpenAI | https://api.openai.com/v1 |
gpt-4o, gpt-4o-mini, o3, o3-mini | Supported |
| Anthropic | Anthropic | https://api.anthropic.com |
claude-sonnet-4-20250514, claude-opus-4-20250514 | Supported |
| OpenRouter | OpenAI | https://openrouter.ai/api/v1 |
200+ models across all providers | Supported |
| Groq | OpenAI | https://api.groq.com/openai/v1 |
llama-3.3-70b, mixtral-8x7b, gemma-3-4b | Supported |
| Together AI | OpenAI | https://api.together.xyz/v1 |
meta-llama/Llama-3-70b, mistralai/Mixtral-8x7B | Supported |
| DeepSeek | OpenAI | https://api.deepseek.com/v1 |
deepseek-chat, deepseek-reasoner | Supported |
| Mistral AI | OpenAI | https://api.mistral.ai/v1 |
mistral-large-latest, mistral-nemo, codestral-latest | Supported |
| ZenMux | OpenAI / Anthropic | https://zenmux.ai/api/v1 |
Multi-provider gateway | Supported |
Local Providers
| Provider | Protocol | Base URL | Description | Status |
|---|---|---|---|---|
| Ollama | OpenAI | http://localhost:11434/v1 |
Run LLMs locally on your machine | Supported |
| LM Studio | OpenAI | http://localhost:1234/v1 |
Local inference server with OpenAI-compatible API | Supported |
Custom Endpoint
You can configure any provider by entering a custom base URL in the settings. The app accepts any endpoint that supports:
- OpenAI-compatible: POST to
/chat/completionswith the standard payload format - Anthropic: POST to
/v1/messageswith the Anthropic Messages API format
For Ollama, the base URL is typically http://localhost:11434 and the model name is the Ollama model ID (e.g., llama3.2).
SSRF protection is enabled by default. Provider URLs are validated via DNS resolution and IP address checks before requests are made. URLs pointing to internal networks (127.0.0.1, 10.0.0.0/8, 192.168.0.0/16, etc.) are blocked.
API Key Management
In the default mode (ALLOW_USER_PROVIDERS=0), API keys are set once on the server via environment variables and never exposed to the browser. Users interact only with your proxy server.
With ALLOW_USER_PROVIDERS=1, each user can enter their own API key and base URL in the settings dialog. Keys are stored in localStorage and never sent to any service other than the configured provider.