Janitor AI proxy with uncensored models: setup in 2 minutes
10 November 2026 · 2 min read
Janitor AI lets you bring your own model through its proxy setting. If the built-in model is not enough, or your current proxy keeps refusing or running into rate limits, you can point Janitor AI at Orrevan: hosted uncensored and abliterated models behind an OpenAI-compatible API.
What you need
- A Janitor AI account
- An Orrevan account with a little credit (top-ups start at $2)
- An API key from Developers
Setup
- Open any chat in Janitor AI and open API Settings.
- Choose Proxy and add a new configuration.
- Fill in the fields:
| Field | Value |
|---|---|
| Configuration name | Orrevan |
| Model name | orrevan/deepseek-v4.1-flash-unc |
| Proxy URL | https://orrevan.com/api/v1/chat/completions |
| API key | your key, it starts with orv_ |
- Save the configuration, then save the settings and refresh the page.
The Proxy URL must be the full address ending in /chat/completions. The model name must be the exact ID, not the display name.
Which model?
| Model ID | Why |
|---|---|
orrevan/deepseek-v4.1-flash-unc | Cheap, fast, 1M token context for long chats |
orrevan/abliterated-model-large-v2 | Our strongest model, best for complex stories |
orrevan/qwen38-obl | Very cheap all-rounder |
The full list with prices is on the homepage. Models behave differently and some filter more than others: if one refuses, switch to an abliterated one.
What does it cost?
You pay per token, no subscription required. A typical roleplay message with about 4,000 tokens of context and a 300 token reply costs around $0.003 on DeepSeek V4.1 Flash Uncensored. So $2 lasts for several hundred messages. Long chats cost more per message, because the whole history is sent each time. Lower the context size in Janitor AI's generation settings if you want to save money.
Troubleshooting
- Network error: check that the Proxy URL ends in
/chat/completionsand has no space at the end. - 401 or "invalid key": copy the key again from Developers. Keys are only shown once when you create them.
- 402 or "no credit": your balance is empty, top up under Credit.
- Model not found: use the exact model ID from the table above.
In Developers you can also give the key a daily spending limit, so a leaked key cannot drain your balance.
Privacy
Orrevan does not store the content of API requests as chats and does not train on your data. For billing we only record token counts. To answer, your input is sent to the provider of the selected model, details are in the privacy policy.
Create an account and get your key.