Janitor AI Not Working? Errors, Slow Replies and Proxy Fixes

Janitor AI can fail in more ways than most chat apps, because it can run on two kinds of model. JLLM is Janitor’s own model. A proxy is an outside model you connect with your own API key, often through OpenRouter. A problem in either, or in your browser, can produce an error that looks similar.

This guide sorts the common errors by where they come from and gives fixes from Janitor AI’s help centre and OpenRouter’s documentation. It is documentation-based: we have not reproduced these errors ourselves. Janitor AI is for adults only, and its help centre says suspected under-18 accounts are permanently banned.

First: JLLM, proxy or you?

The single most useful test comes from Janitor AI’s own help centre: switch the chat to JLLM. If JLLM replies normally, your proxy or its provider is the likely problem. If JLLM fails too, the issue is either Janitor AI itself or your browser and network.

Then check for a wider problem. Third-party trackers such as StatusGator list Janitor AI’s frontend and backend status, and the official subreddit and Discord usually show outage reports quickly. Trackers can lag, so look for more than one sign.

Is it you or them?

Proxy and API key errors

Janitor AI’s OpenRouter Error Guide is the most detailed official source. Before checking any error, it says to make sure your configuration is active, click Save Settings even if nothing changed, fully refresh the page, then reopen the config and click Test.

Common errors, per the help centre

The free-model limits behind many 429s

OpenRouter’s own documentation says free models are limited to 20 requests a minute. The daily limit is 50 requests if you have bought fewer than 10 credits, rising to 1,000 if you have bought at least 10. It also says a negative balance can cause 402 errors even on free models. Janitor AI’s guide adds that failed responses count toward the daily cap, so repeated retries can use it up faster.

Other proxy fixes from the help centre

• Generate a new API key. Keys can break; create a fresh one and test it.
• Check model filters. Some models can fail silently if a prompt triggers their filters. Rephrase or switch models.
• Block a problem provider. The guide describes adding a poorly performing provider to OpenRouter’s ignored providers list.

Site, network and browser errors

Rate Limit or Network Error on the site

Janitor AI’s help article for this error lists proxy problems as the main cause, especially when popular model servers are down or overloaded, plus browser or VPN issues. Its suggested steps:

1. Switch to JLLM to test whether the proxy is the cause. If it is, wait about 24 hours and try again.
2. Clear your browser cache and cookies.
3. Try a private or incognito window.
4. Switch browsers.
5. Toggle Wi-Fi off and on, or Airplane Mode on a phone.
6. Change your VPN setting: turn it off if you use one, or try one if you do not.

Cloudflare “Access Denied”

The help centre says Cloudflare’s security layer flags your connection, not your account, based on things like IP address, browser fingerprint and network reputation. Its fixes include clearing cache, trying another browser, checking that your device clock is correct, disabling privacy extensions or Brave Shields for the site, updating your browser and trying another network. Support cannot override Cloudflare’s decisions.

Error 502

The help centre describes 502 as a temporary server-side issue with the chat system that usually clears on its own. Wait a few minutes and retry. If it lasts a long time, contact support.

Slow responses

Janitor AI’s documentation that we could read does not give official figures for JLLM wait times. Slowness generally reflects load on JLLM or your proxy’s provider. Practical options:

• Try a quieter time of day.
• On a proxy, switch to a less busy model.
• Trim very long chats. The help centre’s token guide says permanent elements and chat history share a budget of roughly 8,000 to 9,000 tokens on JLLM; leaner context can help.

Worked example

You use a free model through OpenRouter. After a busy evening, every message fails with a 429 mentioning free models per day. You retry several times. Under the limits above, each failed attempt may count toward the 50-request cap, so retrying makes it worse. The fix is to switch to JLLM for the rest of the day, or wait for the reset, which the help centre gives as 00:00 UTC.

If nothing works

If Janitor AI keeps failing at the times you use it, compare other roleplay and companion apps. See our Character.AI alternatives or the companion matcher. For related platforms, see Chub AI not working and Character.AI not working.

What we could not confirm

• Whether Janitor AI runs an official public status page.
• Current JLLM wait times or queue behaviour.
• Error behaviour for proxies other than OpenRouter, beyond what the help centre lists.

For how we source and date guides like this, see how AITried reviews AI products.