App Integration

This site is an OpenAI-format-compatible gateway — any client that supports a "custom OpenAI API URL" can connect.

This page first gives the universal connection parameters, then walks through detailed configuration for Cherry Studio, Claude Code, Codex CLI,followed by a quick reference for other clients and universal troubleshooting; the multi-gateway one-click switcher CC Switch has its own page — see CC Switch.

Common Connection Parameters

All a client needs is two values:

ParameterValue
API URL (Base URL)https://api-model.shuntianchuang.cn (some clients require /v1, i.e. https://api-model.shuntianchuang.cn/v1)
API keyA site token sk-your-token (created on the API Keys page; see API Key Management)

Copy model names from Model Plaza (case-sensitive); naming conventions are in Models.

Cherry Studio

Cherry Studio is a multi-provider desktop AI client (Windows / macOS / Linux). Connect via a "custom provider" and you can use every platform model in chat.

Configuration Steps

  1. Open Cherry Studio and go to "Settings" (bottom-left) → "Model Services".
  2. Find "Add Provider" (or "Add Custom Provider") and choose the OpenAI-compatible type. Cherry Studio add provider
  3. Fill in the provider info: Cherry Studio add provider 2
  4. Add models for the provider: Cherry Studio add provider 3
    • If the client offers a "fetch model list / check" capability, click it to pull this site's models automatically
    • You can also add them manually — copy names from Model Plaza (case-sensitive); starting with deepseek-v4-flash is a good test Cherry Studio add provider 4
  5. Return to the chat view, pick the provider and model you just added, and send a test message.

Verifying Success

If you get a model reply, you're connected. You can also see this call's record and cost on the console's "Usage Logs" page — see Billing & Top-up.

Troubleshooting

SymptomCheck
401 / invalid keyKey copied incompletely or disabled — copy it again from the API Keys page
Model list won't loadAdd model names manually; doesn't affect usage
Reply hangs with no outputStreaming is on but the network layer buffers SSE — turn off streaming or check the proxy; see the FAQ
Model unavailableThe model may not be in your token's group — see the groups section of Models

Claude Code

Claude Code is Anthropic's official command-line coding assistant. The site has opened the native Anthropic endpoint (/v1/messages) — point Claude Code's request URL at this site with a site token and you're set, billed by normal usage.

Setting Environment Variables

Two variables are needed: ANTHROPIC_BASE_URL (pointing at this site) and ANTHROPIC_AUTH_TOKEN (a site token). Setting ANTHROPIC_MODEL to pick a default model is also recommended — Claude Code's default claude-* models are not hosted here, and without it you'll get "model not found".

macOS / Linux (bash)

# Add to ~/.zshrc or ~/.bashrc, then source to apply
export ANTHROPIC_BASE_URL="https://api-model.shuntianchuang.cn"
export ANTHROPIC_AUTH_TOKEN="sk-your-token"
export ANTHROPIC_MODEL="glm-5.2"

Windows (PowerShell)

# Applies to the current session; write into $PROFILE to persist
$env:ANTHROPIC_BASE_URL = "https://api-model.shuntianchuang.cn"
$env:ANTHROPIC_AUTH_TOKEN = "sk-your-token"
$env:ANTHROPIC_MODEL = "glm-5.2"

Create the token on the API Keys page; see API Key Management.

Available Models

ANTHROPIC_MODEL accepts models that support the Anthropic endpoint.

Verification

Start it in any directory:

claude

Ask anything (e.g. "introduce yourself in one sentence") — a normal reply means success. You can also confirm the call on the console's "Usage Logs" page.

Troubleshooting

SymptomCheck
model not foundANTHROPIC_MODEL not set, or set to a model not in the list above
401 / invalid keyANTHROPIC_AUTH_TOKEN not applied (open a new terminal or source the config), or token disabled
Insufficient balanceTop up first; also check the token's own quota in Error Codes
Slow responsesCheck line status via "Test latency" on the console Overview's API info card

Codex CLI

Codex CLI is OpenAI's (ChatGPT's) official command-line coding assistant — point its model provider at this site via a config file and you're set.

Configuration Steps

  1. Edit (create if missing) the config file ~/.codex/config.toml:
~/.codex/config.toml
model = "deepseek-v4-flash"
model_provider = "shuntianchuang"

[model_providers.shuntianchuang]
name = "shuntianchuang"
base_url = "https://api-model.shuntianchuang.cn/v1"
env_key = "SHUNTIANCHUANG_API_KEY"
wire_api = "chat"
  1. Set the environment variable (get the key from the API Keys page; see API Key Management):
export SHUNTIANCHUANG_API_KEY=sk-your-token

Write it into ~/.zshrc or ~/.bashrc to persist.

  1. Verify — run in any directory:
codex "Hello"

A reply means you're connected.

Notes

  • model can be any Chat-type model from Model Plaza, e.g. glm-5.2, kimi-k2.6, qwen3.7-plus (model names are case-sensitive)
  • wire_api = "chat" means the OpenAI Chat Completions protocol (/v1/chat/completions), which this site fully supports
  • Models supporting the OpenAI Responses protocol can be filtered by "endpoint type" in Model Plaza; see Models · Multi-Protocol Endpoints

Troubleshooting

SymptomCheck
model not foundMisspelled / wrong-case model name in config.toml — copy from Model Plaza
401 / invalid keySHUNTIANCHUANG_API_KEY not set, or token disabled
Insufficient balanceTop up first; see Error Codes
Config not appliedConfirm the file is ~/.codex/config.toml; reopen the terminal after changes

Troubleshooting

SymptomCheck
Verify failsMissing /v1, or the key has stray spaces; make sure it's https://api-model.shuntianchuang.cn/v1
Model list emptyAdd model names manually; see Models
401 / 403Token disabled, or restricted by model scope / IP allowlist — see Error Codes
Usage unchangedThat feature used Cursor's own models rather than the custom API — check that the selected model is a custom one

Other Clients

ClientPlatformHow to connect
Lobe ChatWeb / desktopLanguage model settings → add an OpenAI provider: API proxy URL https://api-model.shuntianchuang.cn/v1, API Key = token, then add models
DeepChatDesktopModel providers → custom (OpenAI-compatible): API URL https://api-model.shuntianchuang.cn, API Key = token, pick / type in models
OpenCatmacOS / iOSCustom provider: API endpoint https://api-model.shuntianchuang.cn, token as the API Key, add models
AionUIDesktopModel services → add a custom OpenAI-compatible provider: Base URL and token as above, add models
Fluent ReaderBrowser extensionPick a custom OpenAI-compatible interface as the translation service, enter the API URL and token, choose a Chat-type model (e.g. deepseek-v4-flash)
CC SwitchDesktopMulti-provider one-click switcher — see CC Switch
TIP

Client UIs iterate quickly, so menu names may differ slightly from the table — but as long as you find the "custom / OpenAI-compatible provider" entry and fill in the common parameters, you're good.

Three Universal Checks

When any client fails to connect, run these three first:

  1. Is the URL right? Mind whether a /v1 suffix is required — see Base URLs & Auth.
  2. Does the key have stray spaces / newlines? Copy it again from the API Keys page.
  3. Does the client support SSE? Streaming needs both the client and the network layer to support it — otherwise turn the streaming option off.

Still stuck? See Error Codes & Troubleshooting, or report via Contact Us.