Most multi-model AI API gateways make a convenient shortcut: they wrap every upstream model behind a single /v1/chat/completions translation layer.
While that works for basic chat prompts, it starts breaking down in real coding agents and production pipelines:
-
Anthropic Claude relies on native
/v1/messagescontent blocks (thinking,tool_use,tool_result) and prompt cache controls (cache_control). -
OpenAI GPT-6 / xAI Grok increasingly target the Responses API (
/v1/responses), which usesinput,output_text,text.format, and named SSE events likeresponse.output_text.deltarather thanchoices[0].delta. -
Image and video models require dedicated
/v1/images/generations,/v1/images/edits(multipart), and async/v1/videosendpoints—not text completions.
DeepClub is built around a protocol-first gateway model: one account, one API key, and pay-as-you-go billing, while routing each model family through its recommended protocol.
1. Protocol Selection at a Glance
| Model Family | Primary Endpoint | SDK Base URL | Fallback / Notes |
|---|---|---|---|
Claude (claude-opus-5-5, claude-sonnet-5-5, claude-fable-5-1/5-2) |
POST /v1/messages |
https://deepclub.net |
/v1/chat/completions available for legacy clients |
GPT Text & Grok (gpt-6.1-sol, gpt-6-astra, gpt-6-luna, gpt-6.1-terra, grok-4.7) |
POST /v1/responses |
https://deepclub.net/v1 |
/v1/chat/completions available for compatibility |
| DeepSeek, Gemini Text, GLM, Kimi, Qwen | POST /v1/chat/completions |
https://deepclub.net/v1 |
GLM also exposes /v1/responses
|
Images (gpt-image-2*, gemini-3-pro-image, nano-banana-*) |
POST /v1/images/generations, POST /v1/images/edits
|
https://deepclub.net/v1 |
Dedicated Images API |
Videos (Seedance-2.0-Fast/Mini, MiniMax-H3) |
POST /v1/videos + polling |
https://deepclub.net/v1 |
Async task ID polling + content download |
2. Two-Line Migration in Python
Claude via Anthropic SDK (Root URL — no /v1)
import os
from anthropic import Anthropic
client = Anthropic(
api_key=os.environ["DEEPCLUB_API_KEY"],
base_url="https://deepclub.net",
)
msg = client.messages.create(
model="claude-sonnet-5-5",
max_tokens=1024,
messages=[{"role": "user", "content": "Explain SSE event parsing in 3 bullets."}],
)
print(msg.content[0].text)
GPT-6 / Grok via OpenAI Responses SDK (/v1)
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["DEEPCLUB_API_KEY"],
base_url="https://deepclub.net/v1",
)
resp = client.responses.create(
model="gpt-6.1-sol",
input="Compare Responses API and Chat Completions for structured outputs.",
store=False,
)
print(resp.output_text)
3. Coding Agents (Claude Code, Cursor, Cline, Roo Code)
-
Claude Code: Export
ANTHROPIC_BASE_URL=https://deepclub.netandANTHROPIC_API_KEY=<YOUR_KEY>(full setup at Claude Code custom ANTHROPIC_BASE_URL setup). -
Cursor / Cline / Roo Code: Override the OpenAI-compatible Base URL to
https://deepclub.net/v1and use exact model IDs from the live LLM API pricing catalog (see Cursor setup and Cline & Roo Code setup).
4. Billing & Privacy Guarantees
- Pay-as-you-go: Text models bill per million tokens (with separate prompt caching read/write pricing where supported); image and video models bill per successful generation.
- Global & Asian Payment Options: Visa/Mastercard, Alipay, and WeChat Pay—no monthly subscription lock-in.
- Zero Data Retention (ZDR): Prompts and completions are not stored persistently or used for training.
If you are building with AI coding tools or multi-model pipelines, you can inspect live endpoints, machine-readable pricing (/llms.txt), and per-family guides on the DeepClub English API Gateway Portal and Python & Node.js SDK Guide.
Top comments (0)