Janitor AI Login _ Sign Up, JanitorLLM Setup and Fixes

Janitor AI Login _ Sign Up, JanitorLLM Setup and Fixes

Sven Bergstrom Updated · November 20, 2026

Janitor AI Login _ Sign Up, JanitorLLM Setup and Fixes Hero Image

To log in to Janitor AI, go to janitorai.com, tap Login at the top right, and sign in with the email code the site sends you, your password, Google, or Discord. New here? Tap Sign up, verify your email with the six-digit code, set a username, and you can browse characters right away. You only need one more thing before your first message: a model. Pick JanitorLLM Beta in the chat settings and you are chatting for free in about two minutes.

Janitor AI is a character roleplay platform, not a single chatbot. It hosts millions of user-made characters, and the replies come from a language model you choose. That choice is where most beginners get stuck, and where most errors come from. This page covers the account, the model setup, and the fixes, in the order you will hit them.

Quick facts

  • Official site: janitorai.com — the only real login page. Any other domain asking for your Janitor password is fake.
  • Cost: the account is free. JanitorLLM Beta is free with unlimited messages and needs no API key.
  • Paid tier: janitor+ launched on 24 June 2026 at $12.99 a month, with about five times the context and priority messages.
  • Founded: June 2023, by Jan Zoltkowski.
  • App: there is no official Janitor AI app or APK. The platform runs in your browser.
  • Age: the platform is 18+. Users in the UK, Brazil and Australia now pass an age check through k-ID.

Sign up and log in, step by step

Open janitorai.com in your browser and tap Sign up. You can register with an email and password, or use Google or Discord sign-in. Pick a username and handle, confirm you are over 18, and check your inbox for the verification code. That code lands in the spam folder often enough that it is worth checking there before you hit resend three times.

Logging back in works a bit differently than most sites. Sign-in now defaults to a one-time email code: you type your email, get a six-digit code, and you are in. If you set a password and prefer it, tap use password instead on the login screen. If the code never arrives, wait a minute, check spam, and make sure you are using the same email you registered with. A login loop that keeps bouncing you back is usually a browser issue, so try a private window with extensions off before you reset anything.

One thing that confuses new users: you can browse and open characters before any setup, but you cannot send a message until a model is connected. That is not a bug in your account. It is the next step.

JanitorLLM Beta setup

JanitorLLM, often shortened to JLLM, is Janitor's own built-in model. It is free, it needs no key and no card, and it is tuned for roleplay and creative writing. Janitor moved it to a new base architecture on 20 April 2026 and shipped a fix in July 2026 that cut repetitive, looping replies a lot in their own testing.

To turn it on, open any character and tap Chat. Open the menu at the top right of the chat and go to the API or Model Settings. Set the provider to janitor. That is it. Send a short test message. If a reply comes back, your account is fully working and everything else on this page is optional.

JLLM reads roughly 9,000 tokens of context on the free plan, and that number is fixed by your plan. You cannot raise it with a slider. Your character's permanent definition, your persona, and the recent chat all share that budget, which is why very long character cards make the bot forget things sooner.

Adding an API key or a proxy

A proxy on Janitor AI is not a VPN. It is a custom API connection: instead of JLLM, your chats go to an outside model provider, and you pay that provider. People do it for longer memory, stronger writing, or a specific model like Claude or DeepSeek.

The setup follows Janitor's own quickstart pattern:

  1. Create an account with a provider such as OpenRouter or DeepSeek and generate an API key. Treat the key like a password.
  2. Open a chat on janitorai.com and open the chat's API settings from the top-right menu.
  3. Choose the proxy or custom provider option and add a configuration: a name, the exact model name, the proxy URL, and your API key, with no extra spaces.
  4. For OpenRouter, the proxy URL is https://openrouter.ai/api/v1/chat/completions and keys start with sk-or-v1-. Copy the model ID exactly from OpenRouter's model list.
  5. Save, use the built-in connection test if your build shows one, reload if the site asks, and send a test message.

The three fields must all belong to the same provider. A DeepSeek key pasted next to an OpenRouter URL fails every time. OpenAI keys can also be added directly in the API settings, but OpenAI charges per token and enforces its own content rules, which is why many roleplay users route through OpenRouter instead. Never use a random "free proxy" URL shared in a comment or Discord post. Whoever runs it can read your chats and steal your key.

Settings that actually matter

Open a chat, tap the top-right menu, and open Settings. Since the redesign on 19 July 2026, provider choice and response length sit under General, custom prompts and forbidden words sit under Instructions, and temperature and max tokens sit under Generation controls.

Temperature controls how varied replies get. Janitor's own help guidance is to start at 0.6 on JLLM, move in steps of 0.1, and stay under 1.0. JLLM ships at 1, so lowering it is the single quickest quality fix most people can make. Max tokens at 0 means the model chooses the reply length; a low number cuts replies off mid-sentence. Response length offers Standard or Shorter, handy when replies turn into essays. On a proxy, a context size around 16,384 is a sane middle ground. Bigger is not smarter, just slower and more expensive.

Your persona lives in your Profile under the Persona tab. Write 50 to 100 words on who you are in the story: name, look, role, how you talk. You can keep several personas and swap per chat. A short, clear persona beats a page of lore the model will forget anyway.

Common problems and fixes

"Network error" after messages worked yesterday. This is rarely your Wi-Fi. It often hides a billing problem: an empty provider balance comes back as a 402 and Janitor shows it as a plain network error. Check your OpenRouter or DeepSeek balance first. A mismatched key and URL is the other big cause.

API key errors. A 401 means the key is missing, wrong, or pasted with a stray space. Paste only the key, with nothing in front of it, and make sure it is saved under the same configuration the chat is using. Still stuck? Generate a fresh key. A 404 usually means the model ID is wrong or retired, so copy it again from the provider's list. A 429 means a rate limit: wait 10 to 15 minutes, or switch models, and stop hammering regenerate, since failed requests can still eat your allowance.

Empty or repetitive replies. Lower the temperature toward 0.6, start a fresh chat, and trim a bloated character card. Long chats also degrade as the context fills up, so a new chat with a short recap of the story often fixes a bot that suddenly went dull.

Error getting characters, or the site will not load. That one is server-side. Wait it out, and try a private window to rule out a cache problem on your end.

Safety and privacy

Janitor AI allows mature roleplay for adults, and you control it: the NSFW toggle sits in your profile settings, and it stays off until you switch it on. Keep it factual and simple. This is an 18+ platform, full stop.

Protect the basics. Only log in at janitorai.com. Never type your password or API key into a lookalike page, and never share a key in a screenshot or a support post. If a key leaks, revoke it and make a new one. Remember that proxy chats pass through the provider you chose, so that company's privacy rules apply to those messages too.

FAQ

Is Janitor AI free? Yes. The account and JanitorLLM Beta are free with unlimited messages. janitor+ is the optional paid tier at $12.99 a month.

Do I need an API key? No. A key is only for the optional proxy route to outside models.

Is there a Janitor AI app? No official app exists. See our download page for the fake APK warning.

Why does the bot forget things? JLLM's context is about 9,000 tokens on the free plan. Shorter cards and personas leave more room for the story.

Once your login works and JLLM answers, you are set up. Everything after that is taste. Start simple, change one setting at a time, and your chats will stay stable.

S

Sven Bergstrom

Original Post

0 Discussions

No discussions yet. Be the first to start!