# XycAi Developer Docs > XycAi is a unified AI API gateway exposing Claude, GPT, Gemini, GLM, DeepSeek, Kimi, Qwen and 20+ other vendors through one OpenAI-compatible interface. Primary base URL: https://cn.chatapi.app (global: https://jp.chatapi.app, backup: https://apicdn.xycai.us). ## Key Facts - Base URL: `https://cn.chatapi.app` (OpenAI SDK `base_url` = `https://cn.chatapi.app/v1`); alternatives: `https://jp.chatapi.app`, `https://apicdn.xycai.us`, `https://apicdn.xyc.ai` - Auth: `Authorization: Bearer sk-xxxxxxxx` (Claude native uses `x-api-key`; Gemini native uses `x-goog-api-key`) - Keys route by vendor (default), by model, or across all models; the gateway picks the cheapest available pricing group automatically - Error structure: `{"error":{"code":"","message":"... (request id: ...)","type":"new_api_error"}}` ## Get started - [Quickstart](https://docs.xycai.us/index.md): XycAi is a unified AI API gateway — one key, one endpoint, 20+ vendors including Claude, GPT, Gemini, GLM, DeepSeek and Kimi. Send your first request in minutes. - [Endpoints](https://docs.xycai.us/endpoints.md): XycAi's four API endpoints — cn.chatapi.app (China direct, primary), jp.chatapi.app (global CDN), apicdn.xycai.us (backup) and apicdn.xyc.ai (overseas direct) — and how to write the base URL for each protocol. - [API keys & routing](https://docs.xycai.us/keys.md): XycAi API key routing modes (by vendor / by model / all models), how smart routing picks the cheapest available pricing group, the group system (Enterprise / Economy / Workflow / Plan / Unreviewed / Unlimited), quota modes and concurrency limits. - [Authentication](https://docs.xycai.us/authentication.md): How to authenticate with XycAi per protocol — Authorization Bearer for OpenAI-compatible routes, x-api-key for Claude native, x-goog-api-key for Gemini native — plus key format, safety tips and troubleshooting. - [Models](https://docs.xycai.us/models.md): Every model XycAi currently serves — by vendor, protocol, group and billing — generated from the live pricing API. - [Model status & pricing API](https://docs.xycai.us/status.md): Two public, unauthenticated XycAi endpoints — GET /api/provider/status returns the live availability of every model×group (24h success rate plus the latest 30-minute probe), GET /api/provider/pricing returns the CNY price list that matches the model square. - [Errors](https://docs.xycai.us/errors.md): XycAi gateway error response structure, error types, common HTTP status codes, and troubleshooting tips. ## API reference - [Chat Completions](https://docs.xycai.us/chat.md): XycAi's OpenAI-compatible chat completions endpoint /v1/chat/completions: parameters, streaming, multimodal, and tool calling. - [Responses API](https://docs.xycai.us/responses.md): OpenAI Responses API on XycAi: /v1/responses over HTTP and WebSocket, used by Codex CLI / Codex App and the official OpenAI SDK. - [Claude Messages](https://docs.xycai.us/messages.md): Call Claude models using the Anthropic native Messages protocol: POST /v1/messages with x-api-key authentication. - [Image Generation](https://docs.xycai.us/images.md): XycAi image generation and editing endpoints /v1/images/generations and /v1/images/edits, compatible with OpenAI Images. - [Video Generation](https://docs.xycai.us/video.md): XycAi asynchronous video generation: submit a task, poll for status, then download the MP4. Supports Veo, Seedance, Dreamina and more. - [Text-to-Speech](https://docs.xycai.us/speech.md): XycAi text-to-speech endpoint /v1/audio/speech, compatible with OpenAI Audio. Returns WAV/MP3 audio. - [Embeddings](https://docs.xycai.us/embeddings.md): XycAi embeddings endpoint /v1/embeddings, compatible with OpenAI Embeddings. Turn text into vectors for search, clustering, and RAG. ## Dedicated channels - [Gemini Native](https://docs.xycai.us/gemini.md): Call Google Gemini using its native protocol: /v1beta/models/{model}:generateContent with x-goog-api-key authentication. - [Jev Judgment Model](https://docs.xycai.us/jev.md): Call the TypeSafe Jev judgment model: POST /v1/decisions (alias /v1/systemone) with state and questions to get classification, score or yes/no answers. Billed on input tokens; output is free. - [Midjourney](https://docs.xycai.us/midjourney.md): XycAi Midjourney-compatible channel /mj: submit drawing tasks and poll for results. - [Suno Music](https://docs.xycai.us/suno.md): XycAi Suno music generation channel /suno: submit music generation tasks and poll for results. ## Agents & clients - [OpenClaw](https://docs.xycai.us/openclaw.md): OpenClaw is compatible with OpenAI / Anthropic APIs — the setup mirrors similar clients. - [Hermes Agent](https://docs.xycai.us/hermes.md): Hermes Agent is compatible with OpenAI / Anthropic APIs — point the service URL at this platform. - [Claude Code](https://docs.xycai.us/claude-code.md): Claude Code is the official CLI assistant. Point it at this platform with your API key. - [OpenCode](https://docs.xycai.us/opencode.md): Connect OpenCode to XycAi through an OpenAI-compatible provider for terminal-based coding workflows. - [Cursor](https://docs.xycai.us/cursor.md): Use Cursor with a custom OpenAI-compatible endpoint and route model requests through XycAi. - [Codex CLI](https://docs.xycai.us/codex.md): Codex CLI works with the platform key through the OpenAI-compatible endpoint. - [Qwen Office Assistant](https://docs.xycai.us/qwen-office.md): Configure Qwen Office Assistant with a custom OpenAI-compatible provider backed by XycAi models. - [Qwen Code](https://docs.xycai.us/qwen-code.md): Qwen Code is a terminal coding assistant that can use XycAi through the OpenAI-compatible API. - [DeepSeek Harness](https://docs.xycai.us/deepseek-harness.md): Switch Harness to an OpenAI-compatible provider and use DeepSeek models through XycAi. - [QwenPaw](https://docs.xycai.us/qwenpaw.md): QwenPaw supports custom model services and can connect to XycAi over OpenAI-compatible APIs. - [Cherry Studio](https://docs.xycai.us/cherry-studio.md): Connect Cherry Studio to XycAi with a custom OpenAI-compatible service for multi-model chats. - [Chatbox](https://docs.xycai.us/chatbox.md): Chatbox supports custom OpenAI APIs, so you can use the unified XycAi endpoint directly. - [Cline](https://docs.xycai.us/cline.md): Use Cline with an OpenAI-compatible provider to route VS Code coding requests through XycAi. - [Qoder](https://docs.xycai.us/qoder.md): Qoder supports custom OpenAI-compatible model services for coding chat and agent workflows through XycAi. - [Qoder CN (formerly Lingma)](https://docs.xycai.us/qoder-cn.md): Qoder CN (formerly Lingma) can use XycAi through a custom OpenAI-compatible service. - [Kilo CLI](https://docs.xycai.us/kilo-cli.md): Kilo CLI can use XycAi through the OpenAI-compatible endpoint and shell environment variables. - [Postman](https://docs.xycai.us/postman.md): Use Postman to validate the endpoint and request parameters before configuring another client. - [Dify](https://docs.xycai.us/dify.md): Add XycAi to Dify as an OpenAI-API-compatible provider for workflows and applications. - [Gemini CLI](https://docs.xycai.us/gemini-cli.md): Use Gemini CLI to call Google Gemini models from the terminal. - [TRAE SOLO](https://docs.xycai.us/trae.md): TRAE SOLO supports OpenAI-compatible third-party providers. - [Node.js Setup](https://docs.xycai.us/nodejs.md): Most CLI clients (Codex, Claude Code, Gemini CLI) require Node.js. Install an LTS release (18 or 20). - [More tools](https://docs.xycai.us/more-tools.md): Any tool that supports the OpenAI-compatible API can connect to XycAi with the shared parameters below. Every page is also served as HTML at the same path with .html instead of .md. Chinese versions live under /zh/ with identical file names (e.g. https://docs.xycai.us/zh/index.md).