Every model.
One endpoint.
Moflux routes your requests to OpenAI, Anthropic, Gemini, DeepSeek and 40+ more providers through a single OpenAI‑compatible API. Swap models by changing a string — your code, your keys and your billing stay in one place.
Start on your phone. No Codex, API key, or setup steps.
# Mainland China · Alibaba Cloud base_url="https://api-cn.mofluxapi.com/v1"
# Global · Cloudflare base_url="https://mofluxapi.com/v1"
Set up the Codex app in minutes.
No coding needed — install the app, paste one line, enter your key. Works on Mac, Windows and Linux.
Your API key is created for you — just copy, paste, done. Or follow the manual steps below.
Prefer to set it up by hand? Show the manual steps
Download the Codex app
Codex is OpenAI's AI coding assistant — it can build websites, write scripts and automate tasks for you. Download it free, open the file and install it like any other app.
Not sure which Mac? Anything from 2020 or later is Apple Silicon.
Linux (Ubuntu / Debian): there is no desktop app — Codex runs in the terminal instead. Skip this download; the setup line in step 04 will offer to install it for you.
Get your API key
Create an account, then copy your key from the console. It looks like sk-...
Registering is free — new accounts get $10 credit, and GPT models are 0.5× half price for a limited time.
Open the command window
Mac: press ⌘ + Space, type Terminal, press Enter.
Windows: press the ⊞ Win key, type PowerShell, press Enter.
Linux: press Ctrl + Alt + T, or open the Terminal app.
Paste the setup line
Copy the line for your system and paste it into the window, then press Enter. When it asks, paste your API key.
curl -fsSL https://mofluxapi.com/codex.sh | bash
powershell -NoProfile -Command "$p=Join-Path $env:TEMP 'moflux-codex-setup.ps1'; Invoke-WebRequest -UseBasicParsing 'https://mofluxapi.com/codex.ps1' -OutFile $p; Unblock-File -Path $p; & $p"
curl -fsSL https://mofluxapi.com/codex.sh | bash
Typed the wrong key? Run the same line again — it lets you replace it. Script blocked or showing an error? Follow the manual setup guide instead.
Start Codex
Open the Codex app — or type codex in a new command window. That's it — everything now runs through Moflux.
- Script blocked or showing an error — set it up by hand instead.
- Can't see the newest model — how to update and pick it.
- App stuck in English — switch the Codex interface to Chinese.
- 401 / "invalid key" / signed-in confusion — the common fixes.
Change a string, not your stack.
If your code speaks OpenAI, it already speaks Moflux. Point your SDK at one base URL, then reach any provider's models by name.
from openai import OpenAI import os client = OpenAI( base_url="https://mofluxapi.com/v1", api_key=os.environ["MOFLUX_API_KEY"], ) stream = client.chat.completions.create( model="claude-fable-5", # or "gpt-5.2", "gemini-3-pro", ... messages=[{"role": "user", "content": "hello"}], stream=True, ) for chunk in stream: print(chunk.choices[0].delta.content or "", end="")
import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://mofluxapi.com/v1", apiKey: process.env.MOFLUX_API_KEY, }); const stream = await client.chat.completions.create({ model: "claude-fable-5", // or "gpt-5.2", "gemini-3-pro", ... messages: [{ role: "user", content: "hello" }], stream: true, }); for await (const chunk of stream) { process.stdout.write(chunk.choices[0]?.delta?.content ?? ""); }
A gateway that earns its place in the request path.
OpenAI-compatible, and then some
Point any OpenAI SDK at /v1 and it just works. Native Claude and Gemini
endpoints are there when you want provider-specific features.
Smart load balancing
Weighted routing across multiple upstream channels, with automatic retries and failover the moment a provider degrades.
Token-level billing
Every request metered to the token. Pay-as-you-go quota, transparent per-model pricing, real-time spend — no surprises at the end of the month.
Keys you can hand out
Per-key quotas, model allowlists, expiry dates and IP restrictions. Give your team and your apps keys — not liability.
Streaming first
SSE pass-through with minimal gateway overhead. Your tokens arrive as fast as the model emits them.
Full observability
Request logs, latency and per-model spend, live in the console. When something is slow, you'll know which provider — and when.
While the model thinks, save the city.
A tiny built-in shooter: pilot the Giant of Light and zap the kaiju before they reach the gateway. One insert coin, zero microtransactions.
DEFENDER
Kaiju are marching on the gateway. You are the last route online.
Start routing in minutes.
Sign up free, get $10 credit, start routing — GPT models at 0.5× for a limited time.
Stuck? A real person is one message away.
Setup help, top-up questions, model requests — join a group and ask. That's also where new models and announcements land first.
One tap to join — works on phone and desktop. Fastest way to get an answer.
Open the QR code page and scan it with WeChat to join the group.
Support and community chat for QQ users. Search this group number in QQ to join.