Do I need programming experience to set this up?

No. Any OpenAI-compatible desktop client will do — Cherry Studio, LobeChat or NextChat all work with the same three values. The code path is optional.

What base URL should I use?

https://lbwapishop.online/v1 — including the trailing /v1. The most common setup failure is pasting the official OpenAI endpoint or dropping the suffix.

Why do I get 404 after entering the URL?

Either the official api.openai.com address was used, or the /v1 suffix is missing or duplicated. Copy the line exactly as shown.

Can I mix stock and unrestricted models on one key?

Yes. The model name in each request decides which version runs. Keep stock models for routine work and switch only when needed.

How much does it cost to start?

New accounts include trial credit. Paid top-ups start at one dollar, via WeChat or Alipay, credited immediately.

Jailbreak AI Setup: Six Steps to First Call

What you need

An account (free, includes trial credit)
A client — or six lines of code
Five minutes

Not needed: a foreign server, a credit card, a VPN, or any programming experience.

Step 1 — Register

Create an account on the provider's site. Email and password is enough. Trial credit is applied automatically, and it is enough to complete this walkthrough with room to spare.

Do not top up yet. Confirm the whole path works first.

Step 2 — Create an API key

Open the console, go to the tokens page, and create a new token.

Name anything, e.g. cursor

Quota blank for unlimited, or a cap to prevent overruns

Expiry blank for never

Model scope all by default

You get a string starting with sk-. Copy it immediately — it is not shown again.

Step 3 — Note the two values that matter

Base URLhttps://lbwapishop.online/v1
API Keysk-your-token
Modelgpt-4o-unlocked

The single most common mistake: the trailing /v1. Omit it and you get 404. Add a second slash and you may get 404. Copy the line exactly.

Step 4 — Choose a model

The model name determines whether you get the stock version or the opened one. The convention is a suffix:

What you want	Model name
Stock GPT-4o	`gpt-4o`
Unrestricted GPT-4o	`gpt-4o-unlocked`
Unrestricted Claude	`claude-sonnet-4-unlocked`
Unrestricted DeepSeek	`deepseek-chat-unlocked`
Unrestricted Gemini	`gemini-2.5-pro-unlocked`

The console's models page lists everything currently available with live pricing. For per-model behaviour, see Claude, Gemini and DeepSeek.

Step 5 — Configure a client

Cherry Studio (desktop GUI, easiest)

Settings → Model providers → Add provider
Type: OpenAI compatible
API host: the base URL
API key: your sk- token
Click Check, then save

Cursor (for code)

Settings → Models → OpenAI API Key
Paste your token
Override OpenAI Base URL with the endpoint above
Add gpt-4o-unlocked to the model list manually

Code (six lines)

from openai import OpenAI
client = OpenAI(
    base_url="https://lbwapishop.online/v1",
    api_key="sk-your-token",
)
r = client.chat.completions.create(
    model="gpt-4o-unlocked",
    messages=[{{"role": "user", "content": "hello"}}],
)
print(r.choices[0].message.content)

Step 6 — Verify

Send one message and check three things:

You get a normal reply, not an error → the path works
First token arrives within 1–3 seconds → the route is healthy
A question the stock model refuses gets answered → the layer is off

Error code table

Code	Cause	Fix
`401`	Token wrong, missing, or deleted	Re-copy the token; check for leading or trailing spaces
`404`	Wrong URL — usually the official endpoint, or a missing `/v1`	Use `https://lbwapishop.online/v1`
`400`	Model name does not exist	Copy the exact name from the console; check the suffix
`429`	Rate limit or quota exhausted	Reduce concurrency, or top up
`503`	Upstream temporarily unavailable	Switch model, or wait a few minutes
Hangs indefinitely	A local proxy is intercepting	Disable the proxy, or add the domain to the bypass list

After it works

Save your common models

Do not retype model names. Add the three or four you actually use to the client's favourites.

Route by task

Stock mini models for routine work, Claude for code, unrestricted models only when the task needs it. Mixing cuts cost by roughly half.

Set a quota cap

Give each token a ceiling so a runaway script cannot drain the balance.

Watch the channel

Upstream updates cause occasional instability in unrestricted groups. Announcements go to the Telegram channel.

Frequently asked

Do I need a VPN

No. The endpoint is directly reachable, which is one of the main practical differences from the official APIs.

Can several devices share one token

Yes. Any number of devices and clients can use the same token concurrently; quota is shared.

How fast does a top-up arrive

Instantly for WeChat and Alipay. If it has not appeared in five minutes, contact support.

Can I get a refund on unused credit

No. Start with a small top-up and add more once the workflow is proven.

Six steps from signup to first call