According to the description, you can use Codex and Claude. It seems like the first 1,000 people to register get a free year of use.
It looks pretty good—if you’re interested, give it a try, though it seems to be a pretty new platform.
At the moment, it should only support Mac, not Windows.
If you want to give it a shot, you can use this invitation code: MIRA-XDFB-TKHY-95CD
Thanks for sharing, I’ll reply first and read it later.
Do I need to download the macOS or iOS version to register?
Yes, I saw that Windows isn’t supported. I’m using the one for Mac.
There are a total of 1,000 slots. Do you get one just by registering, or is there some catch?
Then, if you select ‘always on full brightness’ in the settings, it should use this account’s quota. Otherwise, it will use up your own Codex and Claude quotas.
I’m not really sure about this. After I downloaded it, I just entered the invite code during registration and it worked.
Thanks, at the moment it’s only available on M-series chips and Android.
Thank you, I used your code to register an account.
After checking it out, this thing is a scam. It doesn’t provide any models itself; it just optimizes your context, so you have to set up your own account. Basically, it means it helps you compress the context locally to save money.
There are tons of similar projects on GitHub—this one just bundles a client together. And then they use this invitation gimmick to grow. Everyone should stay away from this kind of junk.
It works, I can use it just fine.
Latest invitation code: you can use MIRA-RHM1-PFMM-W401
You need to download the client and register before you can use the code.
Yes, it’s just trash.
Next time, make sure you really understand it before sharing, okay? This thing is honestly just a total trap.
By the way, just to add: there really is a model. I can use it on a new computer, but if you’ve configured cc before, it might be using your own key. You can try selecting the full fallback option here and see if that helps.
Try it yourself—just sending ‘hi’ twice, and your quota is gone.
I tried using Sonnet and it works pretty well. The speed is quite fast and it also supports web search. At least it doesn’t just say hi twice.
I just sent a “hi” and asked what model you are, and by the third question, it crashed—already hit the limit…
I tried it too, and it just doesn’t work at all.

