Why is JanitorAI not working?

When JanitorAI stops working, it is usually one of three things: the site itself is down, the outside model you connected through a proxy is down or misconfigured, or your browser or network is being blocked. Check the official status page first, then switch to the free JLLM model to see whether the problem is JanitorAI or your proxy. Everything below comes from JanitorAI’s official help center, as checked on Oct 25, 2026.

Step 1: Check if JanitorAI is down

JanitorAI has an official status page at status.janitorai.com. It shows two components, Backend and Frontend. Both were operational when I checked on Oct 25, 2026.

• If either component shows an issue, the problem is on JanitorAI’s side. Wait it out; nothing you change will help.
• If both are operational, the problem is more likely your proxy, your account session or your network. Keep reading.

Step 2: Switch to JLLM to find the culprit

JLLM (JanitorLLM) is JanitorAI’s free built-in model. It is the default unless you have connected an external API.

This is the most useful test in this guide. Disconnect or switch away from your external API configuration and send a message with JLLM.

• JLLM works: JanitorAI is fine. The problem is your proxy provider or your API settings. Go to Step 3.
• JLLM also fails: the problem is your session, browser or network. Go to Step 4.

Step 3: Fix proxy and API key errors

The help center says that a “rate limit” or “network error” message usually means your proxy provider, such as DeepSeek or Claude, is down or overloaded. That is the provider’s server, not JanitorAI’s.

Check your configuration

According to the official proxy quickstart guide, the setup lives inside a chat:

1. Open a chat and click the gear icon.
2. Go to API Settings > Add Configuration.
3. Enter a configuration name, the model name, the proxy URL and your API key.
4. Click Save, then refresh the page.

The guide’s OpenRouter example uses the model name deepseek/deepseek-chat-v3-0324:free and the proxy URL https://openrouter.ai/api/v1/chat/completions. Typos in either field are an easy way to break things. So is forgetting to refresh after saving.

Check your provider’s limits

The official guide notes that OpenRouter’s free usage was 50 messages a day, resetting at midnight UTC, rising to 1,000 a day after a one-time $10 purchase. Those are OpenRouter’s limits, not JanitorAI’s, and they may change. If your messages stop at the same point every day, this is the first thing to suspect.

Costs and logging also depend on the external provider, so check their own dashboard for usage, credit balance or a revoked key.

Provider-specific errors

The help center has separate articles for errors from specific providers, including OpenRouter and Chutes, and for “upstream invalid response” messages. These come from the provider. Try another model on the same provider, or switch back to JLLM, to confirm.

The no-proxy option

If you are a Janitor+ subscriber on the web, the janitor+ Router gives pay-per-token access to third-party models without setting up an external proxy. Per the official help article on Oct 25, 2026, there is no separate Router fee, and credits are non-refundable.

Step 4: Fix 401, 403, 502 and Cloudflare errors

401 Unauthorized

The help center ties 401 errors to your login session or token. Causes it lists: session or token problems, cookies, browser bugs and a blocked IP. Try, in order:

1. Log out and log back in.
2. Clear cookies for JanitorAI.
3. Try a different browser, or a private window.
4. If you use a VPN, turn it off and try again.

Cloudflare “Access Denied”

JanitorAI sits behind Cloudflare, which can block you before you reach the site. The help center lists these causes:

• A flagged IP address or VPN
• Privacy extensions in your browser
• An outdated browser
• A wrong system clock

The fixes follow from the list: disable the VPN, pause privacy extensions for the site, update your browser and set your device clock to automatic. The clock one sounds odd, but security checks rely on time stamps, so a clock that is off can fail them.

403 and 502

The help center lists 403 and 502 among general site and chat issues. As a rule of thumb, a 403 means a request was refused, so go through the 401 and Cloudflare checks above. A 502 means a server in the chain did not answer properly, which points back to Step 1 (status page) or Step 3 (your proxy provider).

Step 5: Check message length and account age

Two limits catch people out, both listed by the help center:

• Message length: messages over 10,240 characters cause an error. If you paste a long scene, split it in two.
• Account age: the help center lists an account under 30 days old among the causes of errors. If your account is new, some actions may not work yet.

Step 6: If the bot works but forgets things

This one is not an error, but it is the most common “it is broken” feeling. The free tier has a context limit of about 9k tokens. Once a chat runs longer than that, the bot loses earlier details. Janitor+ gives 5x more context, according to the official subscription FAQ on Oct 25, 2026. A connected external model may have its own context size set by the provider.

If it keeps failing

If you have been through the status page, the JLLM test and the fixes above and JanitorAI still will not cooperate, it may be time to try something else for a while. Reviewers on Trustpilot also mention removed features, chats cut off and tags not loading, from a small sample of 35 reviews at the time of checking.

I keep a list of JanitorAI alternatives with free plans and prices, so you can compare without signing up for five things at once. For the full picture of the service itself, see my JanitorAI test drive.