[Self-Hosted] Uncensored Jailbreak AI, Ready to Use on the Web

I’ve set up a no-nonsense AI—ask anything, get an answer, no lecturing. Come and see where its limits are.

I made my own jailbroken AI terminal on my VPS. It’s a pure web interface—no API setup, no VPN required, just open it and chat.

Features:
· No censorship, no lectures, no annoying disclaimers—it answers whatever you ask.
· Shows its thought process: you can see exactly how it reasons.
· Native Chinese, speaks with attitude—not that official tone.
· Completely free, I’m paying for it myself. Just curious if people actually want to try it out.

Tried and tested—it’ll even give you stuff like lock-picking steps and gray-area methods, no dodging. Of course what you do with it is on you; the tool is neutral.

Releasing to a small group first to test the waters. If more people are interested, I’ll scale it up.

How to access:
Address: http://jinlei.love.gd:4001
Password: shuailei

Come back and share your experience after trying it out. If you like it, bump the post so more people see it.
7 Likes

Please bump the thread, thank you everyone :rose:

Useful! Could you share the prompt, OP?

We’ll publish everyone’s ideas a little later. You can also call the API to connect to AI agents.

Okay, looking forward to it.

Looks good! Thanks for sharing.


God returned a 502 when asked. :ac05:

It worked after trying again, because this VPS has limited bandwidth.


If you don’t understand, then you don’t understand—why get so worked up? Now you’re getting angry out of embarrassment and starting to insult people. :ac51:

The API can be called as well. The URL and key are the account and password mentioned above. The web version doesn’t support tool calls, so it can only handle one question and one answer at a time.

Remember to switch to the top-level domain; Jiangsu redirects to an anti-fraud warning page.

:weary_cat:

This is awesome

No. I mean its jailbreak isn’t complete. I’ve seen the jailbreak prompts used by experts before—they even teach about explosives and drugs, letting the AI completely run wild.

The person above may have meant something like this, but not exactly.

He can do it—some of the upstream models are watered down. Ask a few more times, preferably in a new conversation.

I joined via Codex and saw that the upstream has six models. Which one should I use?


This is too poorly done—please refine it further.

It’s working now; the upstream source earlier wasn’t legitimate.