Skip to main content

Using JanitorAI with NanoGPT

A quick guide to connecting JanitorAI chats to NanoGPT’s API.

Setup Instructions

  1. Get your API key from nano-gpt.com/api
  2. In JanitorAI, open API Settings
  3. Select Proxy as the provider
  4. Click + Add under Proxy Configurations
  5. Configure the proxy with the following values:
  6. Click Apply, then make sure the NanoGPT proxy configuration is marked Active
JanitorAI now routes your conversations through NanoGPT.
Use nano-gpt.com here, not api.nano-gpt.com. JanitorAI’s app sends NanoGPT requests through JanitorAI’s own servers, and those can’t reach api.nano-gpt.com, even though it works on the website. The nano-gpt.com address works in both, with the same key, models and prices.

Available Models

auto-model-basic automatically selects a model included in your subscription. To use a specific model instead, copy its exact current model ID from the NanoGPT pricing page.

Include Reasoning

If you want JanitorAI to include reasoning output for reasoning-capable models, duplicate the configuration above and change the proxy URL to:

Troubleshooting

If the JanitorAI app shows PROXY ERROR: No response from bot like this:
JanitorAI proxy error saying no response from bot
check the Proxy URL first. It must start with https://nano-gpt.com, not https://api.nano-gpt.com, because JanitorAI’s app can’t reach api.nano-gpt.com. If JanitorAI shows a PROXY ERROR with a status code, the most common causes are:
  • 401: the API key is wrong or has extra spaces.
  • 402: you used a paid model that is not included in your subscription, and your NanoGPT account has no balance.
  • 404: the model ID is outdated, misspelled, or unavailable. Copy the model’s current ID from the NanoGPT pricing page.
Also re-check that Proxy URL is https://nano-gpt.com/api/v1/chat/completions or https://nano-gpt.com/api/v1thinking/chat/completions. If replies have missing letters or misspelled words, or repeat the same phrase without stopping, check Rep. penalty in JanitorAI’s generation settings. 1 turns it off. Values below 1 make the model repeat itself, and on long chats they can produce garbled text and endless loops. If changing it fixes one chat but not another, check the other chat’s generation settings too.