This is a short tutorial for setting up proxies with OpenRouter. As of July 18, 2025, this is a provider with a true free tier for some pretty good models (ie. DeepSeek). Gemini is the other free option, but its setup is much more complicated. I can make another tutorial if there is demand. :)

Update (July 27, 2025): JanitorAI added proxy profiles. The setup is still similar. You can now create one proxy profile for each model you want to use. Be sure to refresh the page when switching. If it still doesn't connect, slightly edit the API key back and forth, then save.

Free Tier (50 requests per day on select models)

1. Create an account on OpenRouter. https://openrouter.ai

2. Click your Profile > Keys.

3. Click "Create API Key".

4. Give your new key a name, e.g., Janitor. Confirm with "Create".

5. Copy and save your key somewhere on your computer. It will be gone after you click away!

6. In Janitor AI, open a chat. Then open API settings. (Menu > API settings OR Click using janitor/proxy). Pick Proxy.

7. Set the following:

   • Model name (choose Custom): deepseek/deepseek-chat-v3-0324:free. This is an example. It may no longer be available.

     • For more suggestions, see below.

   • Proxy URL: https://openrouter.ai/api/v1/chat/completions

     • Do not click "Add/chat/completions to url". It's already in the URL.

   • API Key: Paste the API key you created on OpenRouter. It is a really long thing that starts probably with "sk-...".

   • This is like an Advanced Prompt for JLLM but with a jailbreak in it.

     • Recommendation: https://cheesey-wizards-organization.gitbook.io/masterlist/prompts-and-troubleshooting/my-prompt

8. Click "Save Settings".

   • (Suggested) Accept the default OpenAI generation settings and then lower the temperature to 0-0.1 (yes!) and max tokens to 0 (infinite).

9. Refresh your Janitor AI chat. Important!

   • If you don't, you will see an error message when you try to chat.

And you are good to go!

Paid Tier

As of July 18, 2025, the pricing is a one-time payment of $10 USD to unlock a higher rate limit of 1000 requests per day permanently for the same free models available to free tier users. If you pick a paid model, there is no limit. However, your account balance will be consumed per use. Each paid model has its own pricing and it is based on token count.

If you can afford it, I suggest the $10 USD. I easily go over the previous 50 limit when I am having some serious fun. Re-rolling a response counts as a separate request.

After you hit your tier limit of 50 or 1000 requests per day, you will start seeing an error message each time you chat.

To load a balance to your account, go to Profile > Credits. Payment is done via credit card or crypto.

OpenRouter Free Models

OpenRouter offers many free-tier models, which change all the time. To see them, click "Models" and change the prompt pricing to 0. Or click this: https://openrouter.ai/models?max_price=0

For the curious, most models are actually not hosted by OpenRouter. They are hosted by providers (like Chutes AI) and presented on OpenRouter. It is like a retail shop.

All models suffer significant quality issues on long context lengths. After 16-20k tokens, I find the DeepSeek models perform at a similar level to JLLM. Here is a study testing LLMs on long-context roleplays! https://fiction.live/stories/Fiction-liveBench-Feb-19-2025/oQdzQvKHw8JyXbN87

If you have any questions, feel free to post and someone might help you out!