API docs
This platform is a multi-protocol AI gateway — use the OpenAI, Anthropic or Google Gemini SDK you already have to call 124+ leading models (Claude / GPT / Gemini / Grok / DeepSeek / Doubao / Kimi / Qwen / MiniMax / GLM / Hunyuan), plus image, video, embedding and knowledge-base APIs. No code changes.
Getting started
SDK guides
Client integrations
Images, video, music and speech from Claude Code, Cursor and Codex
Switch with a single ANTHROPIC_BASE_URL env var
A custom model_provider in config.toml
The GOOGLE_GEMINI_BASE_URL env var
Configuring a custom OpenAI base URL
OpenAI Compatible Provider
OpenAI Compatible provider (no longer maintained)
Configuring several models in config.json
Desktop client: one address per protocol
Custom provider on desktop, mobile and web
In-app settings or self-hosting env vars
Deploy with the BASE_URL env var
Connecting a self-hosted ChatGPT-like UI
A new base URL for ChatOpenAI / ChatAnthropic
The openai/ prefix in the SDK and the proxy
One baseURL with createOpenAICompatible
The official fork — zero configuration
Reference
124+ callable models × pricing × context × capabilities
Model family × the four client protocols: native, translated or unsupported
Request/response schemas for every endpoint, with curl examples
Images per image, video per second, and the embeddings endpoint
Read-only knowledge search over REST, plus MCP
Key format, four auth methods, and how billing works
What 400 / 401 / 402 / 502 mean, and how to debug them
Guides
Parsing SSE, handling interruptions, and when stream=true is required
How Function Tools / Tool Use / Function Calling line up
You only change two lines
From the Anthropic SDK's point of view