People open JannyAI for one thing: the Janitor AI character definitions that Janitor itself hides, readable and downloadable. It is a directory and nothing more: no model, no proxy setting, no API. Below: how to pull a card out of JannyAI, then the three values that run it on an uncensored model, whether in SillyTavern, through Janitor AI's proxy setting or right here in our chat, with the price of 100 messages taken from the live catalog. We read JannyAI's pages on 27 October 2026.
If Janitor AI with a proxy is already your setup, the Janitor AI proxy guide gets you there faster.
JannyAI in one paragraph
JannyAI indexes Janitor AI character cards. The page title says "JannyAI - Former JanitorAI.me". When we read it on 27 October 2026 the home page counted 2,159,082 entries, offered tag filters, printed a token count on every card and had a "Hide Low Quality Bots" toggle. The About page explains the reason for the site: Janitor AI hid bot definitions, so JannyAI publishes "the full bot definition/bot description" so you can download it for SillyTavern and other sites. Every character page carries a Character Definition block and a Download button.
Nothing else is there. No chat box, no model, no proxy field, no API. People searching "janny ai proxy" or "jannyai api" want Janitor AI's proxy setting and an API to point it at. Both halves are below: pulling the card out of JannyAI, then running it.
Getting a card out of JannyAI
- Locate the card. Filter jannyai.com by tag, or take any Janitor AI character link and, as the About page puts it, "replace janitorai.com with jannyai.com".
- On the character page, open Character Definition. That is the personality, scenario, example dialogue and opening message Janitor AI keeps hidden.
- Press Download. You get a file that imports into SillyTavern and any other client that reads Character Card JSON or PNG.
- Note the token count under the card's name. That number is the definition's size, it rides along with every message, and it drives most of the cost worked out below.
For anything JannyAI changes after 27 October 2026, its own pages are the reference, not this guide.
Three settings, any client
Every client below connects with the same three values. The model id is the last one; take it from the price table further down.
- Base URL
- https://api.unrestrictable.ai/v1For SillyTavern and most other clients; keep the /v1
- Proxy URL
- https://api.unrestrictable.ai/v1/chat/completionsFor Janitor AI and other clients with a proxy field
- API key
- sk-unr-live-…Made in the console; shown one time only
- Model
- cydonia-24b-v4.3Any id on the models page, copied exactly
SillyTavern
Import the file you downloaded from the character panel. Then set API to Chat Completion, Chat Completion Source to Custom (OpenAI-compatible), Custom Endpoint to https://api.unrestrictable.ai/v1, paste the key, press Connect and choose the model from the list that appears. Every field and every error is in the SillyTavern guide.
Janitor AI
Stay on Janitor AI and only change what answers. In the chat's API settings pick Proxy, enter https://api.unrestrictable.ai/v1/chat/completions as the URL, then the key, then the model id. Each box is covered in the Janitor AI proxy guide. The card itself never leaves Janitor; you only need JannyAI to read or save the definition.
Our chat, no setup
The roleplay chat runs these same builds in the browser, a few turns free each day, no account needed. To use a JannyAI card there, drop its definition into the system prompt in the playground, which has that field.
Uncensored models that take a Janitor card
Each build here is tagged roleplay in the catalog, has no stock refusal layer, and speaks the OpenAI chat format all three clients use. Prices are pulled from the live catalog at render time.
| Wire formats | |||||
|---|---|---|---|---|---|
| Stheno 8B v3.2 | Roleplay | $0.7035 | $0.7035 | 16K | |
| Cydonia 24B v4.3 | Roleplay | $0.484 | $0.60 | 131K | |
| Uncensored AI Roleplay | Roleplay | $0.60 | $2.34 | 128K | |
| Anubis 70B v1.1 | Roleplay | $1.085 | $1.085 | 32K |
More roleplay tunes: the models page. The daily check verifies that each id really serves the model named on it.
Cost of 100 messages with a JannyAI card
Assumptions: a 3,000-token card (the figure JannyAI's counter gave for typical cards on 27 October 2026), 2,000 tokens of history sent with each message, replies of 300 tokens, 100 messages in all.
| Build | 100 messages | Each message |
|---|---|---|
| Stheno 8B v3.2 | $0.37 | $0.0037 |
| Cydonia 24B v4.3 | $0.26 | $0.0026 |
| Uncensored AI Roleplay | $0.37 | $0.0037 |
| Anubis 70B v1.1 | $0.58 | $0.0058 |
On each message the card alone makes up 60% of the input. Cutting a 3,000-token card down to 1,500 saves more than switching to any other model in the table. Top-ups begin at $10, by card, WeChat Pay or crypto, and credit lasts 12 months; the full rate list is on the pricing page.
Test the key once before chatting
The request below returns the model ids your key can use. It is free, and a 200 carrying a list proves both the key and the URL.
curl https://api.unrestrictable.ai/v1/models \ -H "Authorization: Bearer sk-unr-live-…"401 is a key problem. "No API key was sent." says the field was left blank; "The API key is not valid." means a typo or a truncated paste. 404 model_not_found is a misspelled model id. 402 means the balance is empty, and nothing is charged for that request. These same three settings, for every other client, are on the docs page.
Questions
No. You chat on Janitor AI. JannyAI is an unrelated directory that displays and downloads Janitor AI bot definitions, built, by its own About page, because Janitor AI stopped showing them (read 27 October 2026).
No. There is no chat on JannyAI. Take the downloaded card to SillyTavern, to Janitor AI through its proxy setting, or drop the definition into our playground. See the playground.
Neither. The proxy field belongs to Janitor AI, and this API is what fills it: https://api.unrestrictable.ai/v1/chat/completions, your key, and a model id. See the Janitor AI proxy guide.
For reading definitions, Chub and the SillyTavern card sites share the card format. To run one on a model you pick, any client with an OpenAI-compatible base URL field, pointed at this API. See Character AI alternatives.
No. The API hosts the model; your client sends only the card and the conversation. A GPU is for people running the weights themselves.
Every roleplay-tagged build runs a Janitor card unchanged. Begin with the 24B tune in the table, then step up to the 70B one for longer and more consistent replies. See the model list.
Every model above is on the uncensored model list with its live price per million tokens. Per-token pricing sets the whole catalog side by side, and the family directory groups the builds by base model.