Janitor AI proxy, uncensored.

One proxy URL, one key, 369 uncensored models. Set it up in two minutes, and fix 429, 401 and 502 when they show.

Set up the Janitor AI proxy

  1. Make a key in the console and top up from $10. The key starts with sk-unr- and is shown once.
  2. In Janitor, open any chat, click the gear at the top right and choose API Settings, then Proxy.
  3. Press Add Configuration and give it any name.
  4. Model Name: manta-flash-1.0, or any id from the models page, spelled the same way.
  5. Proxy URL: https://api.unrestrictable.ai/v1/chat/completions.
  6. API Key: your key, with no spaces around it. Leave Custom Prompt blank.
  7. Press Add, then Save Settings. Reload Janitor and send a short test line.

Which model makes the best roleplay proxy

The one that holds your character over a long chat at a price you can live with. Start with manta-flash-1.0 for speed and a long context, or pick a roleplay tune from the roleplay page. Every model has its price per million tokens and its context window on the models page; set Janitor’s context size no higher than that window.

Proxy error 429 on Janitor AI

Cause: one of two things. Your key sent more requests in a minute than its limit allows (a new account has a lower limit for its first 24 hours), or the model you picked is busy right now.

Fix: wait a few seconds and send again. Turn off any extension that resends on failure, since retries in a loop keep you over the limit. If the model stays busy, switch the Model Name to another id. A busy-model 429 is never charged.

Proxy error 401 on Janitor AI

Cause: the key is not valid. It was pasted with a space or a missing character, it was revoked in the console, or it belongs to another service.

Fix: copy the key again from the console (make a new one if you lost it) and paste it into API Key with nothing around it. It must start with sk-unr-. Save Settings and reload Janitor before you test.

Proxy error 502 on Janitor AI

Cause: the model did not answer in time, its answer was cut off, or that model is not available right now. A long chat with a huge context makes a timeout more likely.

Fix: send again. If it repeats, switch to another model id, or lower the context size and the max reply length in Janitor. Nothing is charged for a request that fails before the model replies.

Other answers you may see

402 means the balance is used up or the key reached its spend limit: top up, or raise the limit in the console. An error that repeats the model name back to you means a typo in the id. A 404 usually means the Proxy URL is missing /chat/completions.

FAQs

Questions.

https://api.unrestrictable.ai/v1/chat/completions. Janitor wants the full chat completions path, not the base URL that ends in /v1.

Any id from the models page, spelled exactly. manta-flash-1.0 is a good first pick: fast, long context and low priced.

Too many requests in a minute for your key, or the model is busy right now. Wait a few seconds and send again; a busy-model 429 is not charged.

The key is wrong or revoked. Paste it again with no spaces around it; it starts with sk-unr-.

The model did not answer, or its answer was cut off. Send again, or switch to another model id. Nothing is charged for a request that fails before the model replies.

The browser chat gives three free turns with no account. The proxy runs on a key, which needs a balance: top up from $10, priced per token.

Try a model before you connect it.

Three free turns in the browser, no account.

Join on TelegramHelp