yesterday, at the Codex hackathon, i ended up solving this problem by building an MCP connector app that turns ChatGPT into Codex, sort of. 1. you get to use GPT-5.5 Pro by exposing your local dev environment to ChatGPT, so you can use 5.5 pro for planning and handover... to codex to execute it 2. you get to 2x your rate limits, just like last month, because apparently ChatGPT and Codex have separate rate limits here is the quick demoshow more

i'm going to sleep now, will launch it tomorrow until then keep using codex on 1x rate limits 😅

I hope this blew ups like crazy and reaches each and every codex user

great use of freewill!

I hope i don’t became reason you guys change this rate limits haha

We already have oracle from @steipete

@steipete Its browser automation based approach this is bit different through local MCP connector app expose through tunnel

It’s hard to make something simple and tasteful, but I promise it’ll be very easy to use

lessgooo....I'm waiting for you to wake up I am shipping like crazy this week and about to reach weekly limit, this looks so helpful

you are gonna love it

enjoy guys! keep experimenting keep tokenmaxxing keep finding the model behaviours which are new

Haha I built something similar: though I didn’t want to publicize it because I don’t want them to take it away. 😅

I started building it as remote coding agent expose through MCP but don’t want infra overhead so trimmed it down to this local tunnel version And it just end up 2xing my rate limits

@brianakaka does pro on GPT have a limit I haven't reached it? I thought you had unlimited chats and only codex was rate limited

@brianakaka yup i've never reached limit on chatgpt as well

keep codex maxxing

exactly 🔥

Can't wait to use it.

working on packaging it as npx command

This lines up really well with a reply I made a couple of weeks ago: At the time I was thinking about ChatGPT and Codex being able to reference each other more fluidly — using Codex to execute tasks and handle the actual work, while letting models inside ChatGPT (even GPT Pro) focus on planning and high-level direction. Your MCP connector is a practical implementation of exactly that idea. It wouldn’t surprise me if something like this ends up getting built in natively once Codex and ChatGPT merge into a single product (possibly as early as this or next Thursday’s update).

That was exactly the idea. At first, I thought about providing a remote sandbox, but I dropped it because of the hackathon time limit and scoped it down to run locally.

this is sick. went a step further and ran the actual codex binary on the chatgpt plan, real agent + /status, no connector in the middle. one command: npm i -g @mundy/cyrus && cyrus

I made something similar

I did something similar making a cloudflare tunnel to a Serena server on my laptop You did the ui better though

thanks i know where i’m better at😅

Haha thats a coincidence, just started doing this myself

you don’t have to built and waste your token for it, you have me lol

I use pi i have a custom harness and setup its best i have a custom mcp as well

btw i've used pi under the hood for this

but does not support custom mcps tho

theres an extension for it

yeah ik but i don't think we want mcp inside mcp like nested tools exposed to chatgpt

i mean they have already built it right their own MCP tooling so why we even need it in the first place

In some ways, you'rethe one who released GPT5.6 🌚 bro

could've used to record 😔😕

Linux support?

yeah, it's completely browser based, u can use it in all chromium based browser's

Couple this with the codex hot mic workaround and it might become a good AIO.

sounds like something they might ban...

This is litt⚡️

Interesting. The bigger signal might be specialization: GPT-5.5 for planning and context. Codex for execution and iteration. The workflow matters more than the individual model.

super interesting idea, in the era where tokens are more and more expensive, looking forward to see & try it (from a person only use codex and rarely use chatgpt, this can be a huge boost)

tomorrow morning you will be able to try it

Been doing this for some time... Don't expose the tricks lol, you'll get it patched They've already started labellinh certain commands as unsafe, and I had to add more tools for read write etc

reeellleeeaaassseee.... 😅 This is awesome!

Can't wait to use it

technically what is this? a chatgpt app ?

MCP app yes, sorry i missed this cause of this was my first semi viral tweet and it was notification hell i was going through

fire ser

soon will be served at your doorstep sir

@jxnlco I was literally just asking my Hermes agent to research for something like this. I needed a replacement for fable and I never use Pro because it’s stuck behind the web interface. Thanks goat!

@jxnlco Launch soon stay tuned :) We can enjoy till oai merge chatgpt and codex brand together

That's pretty cool I made a plugin where codex agents can defer difficult questions to pro in a token optimized fashion But this is arguably the more elegant approach

Would love to get access.

Where can I go to set this up and use it?

Sharing this soon

I'm guilty of using chatgpt app to do quick changes to my repos, now it's gonna be even better

smart workaround given how pricey tokens are getting. as a mostly-Codex user who barely touches ChatGPT, doubling rate limits this way sounds like a real win — want to try it myself

Let me know how to get it for tests

Soon

is that your dog ?

naah street dog lol it was late night 3 AM when i recorded it