The only thing sent over the wire is the encrypted chat messages e2e. Here's an architectural diagram: github.com/agentify-sh/chat/…
Another feature it has is ability to publish your chat session with redaction so you can share with your team.
Would love some feedback!
chat.agentify.sh is a remote control for codex/grok/claude cli and dream goal is to become a universal remote that will let you use all of them from a single chat. Everything lives on your browser including the chat.
Replying to @thsottiaux
should've reset before the grok promo thing now I tasted grok I can see codex has serious competition
Grok 4.6 xhigh is impressive. Reminds me of how excited I felt when codex first came out. Fast, token efficient & quite capable at completing tasks.
Not sure I'd pay $300/month after promo expires but as codex $200/month user but @elonmusk this has legs keep it up!
built something similar here many months ago
github.com/agentify-sh/safee…
Replying to @mattshumer_
had this happen way back months ago which I solved by building this
github.com/agentify-sh/safee…
Replying to @tom_doerr
something that can do isometric sprites consistently would be amazing
3.5 pro beats fable 5 and gpt 5.6 ?
great can we actually use the fucking thing without weekly usage limits being drained in just a few prompts ?
Replying to @jeremyphoward
how come I never get invited to anything
I been spending so much money on codex and API
guys be very careful when installing agent skills
don't trust but verify carefully
matter of fact just run git tea locally no internet
Replying to @warmademe
its good but I think it still has a long way to go to challenge gpt 5.5
Replying to @JackWoth98
Thanks for the update. Would be great if you could match the usage limits from Codex, I really like Gemini but it's tough to use it for extensive sessions that the same $200/month from Codex. Also it would be great if there was a AG CLI plan only, nothing else, and more usage.
Replying to @JustBill1182 @kimmonismus
I've banned the usage of 5.5-medium because of the sloppiness it introduces to code. I am using only xhigh am but even it is making mistakes
I wonder if they are silently routing
