Skip to content

Provider Setup

This section guides you through configuring various LLM providers in CCX.

General Workflow

  1. Log in to the CCX admin console (default http://localhost:3000)
  2. Select the appropriate proxy endpoint (Chat, Messages, etc.)
  3. Click "Add Channel"
  4. Fill in the channel configuration
  5. Save and test

Key Configuration Fields

FieldDescription
NameDisplay name for the channel
Service TypeUpstream API protocol: openai, claude, gemini, responses
Base URLUpstream API endpoint
API KeysAuthentication keys (supports multi-key rotation)
Supported ModelsModel allowlist for this channel
Model MappingMap request model names to upstream model names
PriorityLower number = higher priority

Service Type Guide

ProviderChat EndpointMessages EndpointNotes
DeepSeekopenai / https://api.deepseek.comclaude / https://api.deepseek.com/anthropicBoth protocols
GLM (Zhipu)openai / https://open.bigmodel.cn/api/paas/v4claude / https://open.bigmodel.cn/api/anthropicBoth protocols
MiniMaxopenai / https://api.minimax.io/v1claude / https://api.minimax.io/anthropicBoth protocols
Kimiopenai / https://api.moonshot.ai/v1claude / https://api.kimi.com/coding/Coding endpoint
OpenAI GPTopenai / https://api.openai.com/v1OpenAI only
Xiaomi MiMoopenai / https://api.siliconflow.cn/v1Via SiliconFlow
Claudeclaude (translation)claude / https://api.anthropic.comNative Messages
Geminiopenai or geminiOpenAI-compat and native

TIP

Most Chinese LLM providers now support both OpenAI Chat and Anthropic Messages protocols. If you use Claude Code CLI, you can configure the Anthropic-compatible endpoint directly under the Messages endpoint.

If you want to configure by client, see the Client Setup guide.

Provider Guides