Janitor AI connects to Anoman through its Proxy API setting — a browser chat app backed by the full guardrail pipeline. The same three values work in any OpenAI-compatible chat app.

Prerequisites

An Anoman API key — create one in the dashboard under API Keys (format anm-sk-…).
A model ID copied exactly as shown on the dashboard Models page, e.g. deepseek-v4-flash-openrouter.
Balance for paid models (top up in the dashboard), or a model from the dashboard Free Models page.

Enable streaming and pick a fast model

Long roleplay prompts on a slow model can exceed the 100-second edge timeout and fail with a 524 error (Janitor may show it as "Failed to fetch"). Turn streaming on so tokens arrive as they are generated and the connection stays alive, and prefer a fast model for long chats.

More detail in the streaming guide.

Add Anoman as a Janitor AI proxy

Open a chat, then go to API Settings → Proxy tab → Custom / "Add Configuration". Janitor wants the full chat-completions URL, not the base URL.

Proxy URL:   https://api.anoman.io/v1/chat/completions
Model name:  deepseek-v4-flash-openrouter
API Key:     anm-sk-your-key-here

Click Test, then Save Settings. If Janitor asks whether to apply OpenAI default generation settings, Yes is fine. Refresh the page after saving.

Some Janitor characters disallow outside proxies. If the proxy option is greyed out, pick another character.

Other OpenAI-compatible chat apps

Chat apps such as SillyTavern, Agnai, or RisuAI that offer an OpenAI-compatible or custom endpoint option use the same values. Most of them ask for the base URL and add /chat/completions themselves; only enter the full URL when the app asks for the complete endpoint, as Janitor does.

Base URL:  https://api.anoman.io/v1
API Key:   anm-sk-your-key-here
Model:     an ID from the dashboard Models page

api.anoman.io accepts browser (CORS) calls on /v1 from any origin, so browser-based chat apps can call Anoman directly.

Fix common errors

Error	Meaning	Fix
`401 auth_invalid`	The key is wrong or incomplete	Copy the key again from API Keys — no spaces, no `Bearer` prefix
`402`	Balance or free allowance used up	Top up in the dashboard, or switch to a model on the Free Models page
`model_not_found`	The model ID does not match the catalog	Copy the ID exactly from the Models page
`403` guardrail code (e.g. `content_violation`)	A guardrail blocked the message	See Errors and Guardrails
`524` / "Failed to fetch"	The reply took longer than 100 seconds	Turn streaming on and use a faster model

Verify the guardrails are active

Every Anoman request is recorded in your dashboard traces with its guardrail results — PII masking and content moderation apply automatically, and prompt injection is scored when capacity allows. Open Traces after a test message to confirm Janitor requests are reaching Anoman.

Need help with your integration?

Reach the Anoman team directly with questions about your setup.

Connect Janitor AI to Anoman AI