How To Use Janitor Ai

Ai
1

I’m trying to learn how to use Janitor AI, but I got stuck setting everything up and I’m not sure what steps I’m missing. I need help understanding how to start, connect the right settings, and actually use the features without running into errors.

2

Most people get stuck on the same 3 parts. Account, API, and model settings.

  1. Make your Janitor AI account.
  2. Open Settings.
  3. Pick an API provider. Usually OpenAI, KoboldAI, or another supported one.
  4. Paste your API key in the right box.
  5. Save it.

If you use OpenAI, you need an active OpenAI account with billing set up. No billing, no replies. That trips up a lot of ppl.

Then do this.

  1. Go to Characters.
  2. Pick a bot.
  3. Check which model it expects, if the bot page mentions one.
  4. Start a chat.
  5. If replies fail, go back to Settings and test a different model.

Important stuff to check if it still doesent work:

  • API key pasted right, no extra spaces
  • Correct API endpoint selected
  • Account has credits or billing
  • NSFW filter/settings match what you want
  • Chat memory or token limits are not set too low

Simple example:
If OpenAI is selected but your key is invalid, Janitor AI usually throws an error or the bot stays stuck loading. If Kobold is selected but no local endpoint is running, same issue.

If you want the easiest route, use one supported provider, add the key, save, pick one bot, and test with a short message like “hi”. Change one setting at a time so you know what broke.

3

@chasseurdetoiles covered the basic hookup part, but honestly the step people miss after setup is that Janitor AI itself is mostly a frontend. It does not magically ‘be the AI’ on its own. You’re basically connecting a chatbot site to a model provider, then tweaking how the convo behaves.

A few extra things to check:

  • Make sure the bot profile has the kind of prompt/style you actually want. Some bots are badly written, so the weird replies are not always your settings.
  • Try a fresh chat after changing major settings. Old chats can keep bad context and make you think setup is broken.
  • Look for generation settings like temperature, max tokens, and memory. Too much creativity = nonsense. Too little = dry robot text.
  • If messages send but replies are awful or short, that’s usually model choice, token settings, or the character prompt. Not always the API.

I kinda disagree with “just test different models” right away. If you change 5 things at once, you’ll confuse yourself fast lol. Better to lock one provider first, confirm it replies, then tune stuff one by one.

Also, if you’re on mobile, some menus are weirdly hidden or janky. Desktop is way easier for first setup tbh. If you say which provider you picked, ppl can probly tell you the exact setting you missed.

4

Big thing I’d add to what @chasseurdetoiles said: separate “account problem” from “model problem” from “prompt problem.”

A clean way to learn How To Use Janitor AI is:

  1. Confirm your API/provider key is accepted.
  2. Use a simple test character, not a heavily customized one.
  3. Send basic prompts like “hi” or “describe yourself in one paragraph.”
  4. Only after that, turn on personas, lore, memory, or advanced formatting.

I slightly disagree on desktop being required. Desktop is easier, yeah, but mobile works fine if you already know where settings live. For first-time setup though, desktop saves headaches.

What trips people up most:

  • Wrong API selected
  • Empty credits/balance on provider side
  • NSFW or model restrictions causing silent weirdness
  • Broken imported character definitions
  • Expecting old chats to “fix themselves” after settings changes

Pros of Janitor AI:

  • Easy frontend
  • Lots of character options
  • Good for quick testing

Cons:

  • Setup can feel confusing
  • Depends heavily on third-party model providers
  • Bad bot configs can make the whole thing seem broken

If you post which provider you chose and what exact error/reply you get, it’s way easier to pinpoint.