Use one Base URL to reach Chinese AI models.

Hermes Agent

Point the model block of ~/.hermes/config.yaml at the gateway, or run hermes model and pick Custom endpoint interactively. base_url carries the /v1 suffix; Hermes appends /chat/completions itself.

model:
  default: kimi-k2.7-code
  provider: custom
  base_url: https://api.chinaapi.ai/v1
  api_key: <your ChinaAPI key>

Claude Code

Export the three environment variables and start claude as usual. The ANTHROPIC_DEFAULT_*_MODEL variables map Claude Code's internal model tiers onto ChinaAPI models, which is what lets you run Chinese models inside an unmodified Claude Code.

export ANTHROPIC_BASE_URL=https://api.chinaapi.ai
export ANTHROPIC_AUTH_TOKEN=$CHINAAPI_KEY
export ANTHROPIC_DEFAULT_OPUS_MODEL=kimi-k3
export ANTHROPIC_DEFAULT_SONNET_MODEL=kimi-k2.7-code
export ANTHROPIC_DEFAULT_HAIKU_MODEL=glm-5-turbo

claude

Codex

Codex talks the Responses protocol, so wire_api must be responses; leaving it on the chat wire format breaks tool calls. Put your key in the environment variable named by env_key.

model = "deepseek-v4-flash"
model_provider = "chinaapi"

[model_providers.chinaapi]
name = "ChinaAPI"
base_url = "https://api.chinaapi.ai/v1"
wire_api = "responses"
env_key = "CHINAAPI_KEY"

Cline, Roo Code, Kilo Code

All three use the same OpenAI Compatible provider form, field for field. Enter the model ID exactly as it appears in Console → Models; these tools do not fetch a model list for custom providers.

Base URL   https://api.chinaapi.ai/v1
API Key    <your ChinaAPI key>
Model ID   glm-5.2

Cursor

Override the base URL under Settings → Models → OpenAI API Key. Add the ChinaAPI model name with Add model first, then leave only ChinaAPI models enabled, because Cursor otherwise sends its built-in model names to your override URL.

Base URL   https://api.chinaapi.ai/v1
API Key    <your ChinaAPI key>
Model      deepseek-v4-pro

OpenClaw

Declare ChinaAPI as a provider with "api": "anthropic-messages" and list the models you want selectable in the session picker.

{
  "models": {
    "providers": {
      "chinaapi": {
        "baseUrl": "https://api.chinaapi.ai",
        "apiKey": "<your ChinaAPI key>",
        "api": "anthropic-messages",
        "models": ["kimi-k3", "glm-5.2"]
      }
    }
  }
}

Aider

Aider routes by model prefix, so keep the openai/ prefix on the model name even though the model itself is a Chinese one.

export OPENAI_API_BASE=https://api.chinaapi.ai/v1
export OPENAI_API_KEY=$CHINAAPI_KEY

aider --model openai/deepseek-v4-pro

Image generation

Send a JSON prompt. Read the generated asset from data[0].url, or from data[0].b64_json when the selected model returns base64.

Key parameters: prompt and model are required. Use size for resolution, n for image count, and response_format for url or b64_json. quality and style are model-specific; common values include standard, hd, or auto.

curl https://api.chinaapi.ai/v1/images/generations \
  -H "Authorization: Bearer $CHINAAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedream-4-5-251128",
    "prompt": "A cinematic skyline at blue hour",
    "size": "1024x1024",
    "response_format": "url",
    "n": 1
  }'

Gemini image generation

Gemini image models use the OpenAI Chat Completions request shape. Read the generated Markdown image from choices[0].message.content; the image payload is a data:image/... URL.

Key parameters: send the signed-in catalog model ID in model and the image prompt in messages. Do not send this model to /v1/images/generations.

curl https://api.chinaapi.ai/v1/chat/completions \
  -H "Authorization: Bearer $CHINAAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<GEMINI_IMAGE_MODEL>",
    "messages": [
      {
        "role": "user",
        "content": "Generate a cinematic skyline at blue hour"
      }
    ],
    "stream": false
  }'

Speech to text

Upload audio as multipart form data; do not set the Content-Type header manually. The recognized text is returned in the text field.

Key parameters: send model and file as multipart fields. Use response_format as json, text, or verbose_json when supported by the selected model.

curl https://api.chinaapi.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer $CHINAAPI_KEY" \
  -F "model=qwen3-asr-flash" \
  -F "file=@speech.wav" \
  -F "response_format=json"

Text to speech

The response body is binary audio, so write it to a file. alloy selects the model's default voice; a model may also publish provider-specific voice IDs.

Key parameters: use input, model, and voice. Select an output with response_format such as mp3 or wav; speed and instructions are available when the selected model supports them.

curl https://api.chinaapi.ai/v1/audio/speech \
  -H "Authorization: Bearer $CHINAAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mimo-v2.5-tts",
    "input": "Hello from ChinaAPI.",
    "voice": "alloy",
    "response_format": "mp3"
  }' \
  --output speech.mp3

Video generation

Video generation is asynchronous. Submit once, save the returned task_id, then poll GET /v1/video/generations/{task_id} until the task succeeds or fails. Creating a task can consume quota.

Key parameters: prompt and model start the task. Use duration, width, height, fps, and n where supported. Resolution is the common clarity control. Put provider-specific controls such as quality, quality_level, negative_prompt, or camera settings in metadata only when they are listed for that model.

curl https://api.chinaapi.ai/v1/video/generations \
  -H "Authorization: Bearer $CHINAAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.1-t2v",
    "prompt": "A paper boat crossing a moonlit lake",
    "duration": 5,
    "width": 1280,
    "height": 720
  }'

curl https://api.chinaapi.ai/v1/video/generations/$TASK_ID \
  -H "Authorization: Bearer $CHINAAPI_KEY"

Embeddings

Send one string or an array of strings. The vectors are returned in data[].embedding in the same order as the inputs.

Key parameters: input accepts one string or a batch. Use dimensions and encoding_format only when supported by the selected embedding model.

curl https://api.chinaapi.ai/v1/embeddings \
  -H "Authorization: Bearer $CHINAAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "text-embedding-v4",
    "input": ["ChinaAPI connects Chinese AI models."]
  }'

Rerank

Rank candidate documents against a query. Read the ordered matches and relevance scores from results.

Key parameters: send a query and documents; use top_n to limit the returned ranked candidates.

curl https://api.chinaapi.ai/v1/rerank \
  -H "Authorization: Bearer $CHINAAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gte-rerank-v2",
    "query": "How do I call Chinese AI models?",
    "documents": [
      "Use one ChinaAPI Base URL and API key.",
      "Install a local database."
    ],
    "top_n": 2
  }'