Yes, you read that right. Keep reading to learn how.
It’s been a minute since I’ve written to this newsletter.
That’s because I’ve been quietly working on a project that’s going to be a game changer for everyone building with AI.
First, some context: AI agents like ChatGPT Codex and Claude Code have taken the internet by storm.
They let you use these models for tasks on your computer like coding, design, and more.
Most importantly, the subscriptions to these agents are heavily discounted compared to normal API rates.
Put differently, the tokens you’re paying for through the API are overpriced compared to what you could get from the subscription.
Although there aren’t official numbers, some estimates say the $200 Codex subscription can deliver up to $14,000 of usage, if that usage were billed at API rates.
That’s a 70x discount.
The problem is: it’s difficult to use your subscription inside of workflows. n8n doesn’t support it, and Vercel serverless functions can’t use it either.
By default you're stuck running it on your own machine or duct-taping a solution together. That’s not something you want in production.
That’s where ProxyLLM comes in. It bridges the gap by routing your subscription through an OpenAI-compatible API.
By connecting your Codex subscription through ProxyLLM, you can use it everywhere you already make API requests.
The only difference is the token cost.
All you have to do is swap the URL. That’s it.
So who’s this for?
AI agencies who are profitable but spending too much time optimizing their AI costs,
Agencies who are barely/not profitable due to their AI costs,
And beginners who should be spending their time landing new clients, not watching token usage.
That last one is the least obvious, but most needed.
When you’re starting out, you should be trying new offers, reaching out to clients, doing anything you can to gain momentum.
Cost-cutting is the least of your concerns, but with token-based billing, it’s hard to ignore.
That’s why it’s crucial to make the switch to ProxyLLM so you can use your subscription instead.
Now you might be wondering, “is this officially supported?”
And I can confidently say: Yes, programmatic use of Codex is officially documented by OpenAI. ProxyLLM simply makes that usage available wherever you already make API calls.
Here’s how to get started:
Visit proxyllm.ai and click “Get Started”
Subscribe to the membership.
Then click on “Codex Hosted” and follow the setup instructions. Takes less than 5 minutes.
If you have any questions, click on “Support” to open a ticket and a real human will get back to you.
There’s a 7-day money-back guarantee so you can try out the whole system risk-free. Just contact support in that time to get a full refund, no questions asked.
Worst case scenario, you don’t like it, and you’re only out a few minutes of setup time.
Best case scenario, you’re able to start saving so you can focus on work that scales the business.
Start scaling without worry: proxyllm.ai
I’ll see you inside.
Yours truly,
Eric Otten
PS. Stay tuned for more updates. We’re building out support for Claude Code subscriptions as well as other features to help you scale without worry!
