DeepSeek API Key in Janitor AI: Setup With Our API

Short answer: what to enter in Janitor AI

If you are wondering how to use deepseek api key in janitor ai, the first thing to settle is which key you hold. A key issued by DeepSeek belongs to DeepSeek's own platform and goes with their endpoint. The key you get from this site is a different key for a different service: an independent, OpenAI-compatible API running one open-weight model, tuned to answer without content refusals for lawful adult use. It is not DeepSeek's official API, and the two keys are not interchangeable.

For this site, the three values Janitor AI needs are:

• API URL: https://api.deepseekapikey.com/v1/chat/completions (the base URL plus the chat completions path)
• API key: the key from your account
• Model name: uncensored

Janitor AI's interface changes from time to time, so field names and labels may differ slightly from what you see. The idea stays the same: choose the option for a custom or proxy OpenAI-compatible API, then fill in those three values.

Step 1: get a key and test it outside Janitor

Create an account on the API key page. Sign-up works with Google or with email and password. New users can claim $0.50 of trial credit with one click on that page, valid for 7 days, one per person. You can start testing without paying anything. Trial keys run at 2 concurrent requests and 60 requests per minute, which is plenty for one person chatting.

Before you touch Janitor AI, check the key from a terminal. This takes a minute and saves an hour of guessing, because it separates "the API is rejecting me" from "Janitor is configured wrong."

Copycurl https://api.deepseekapikey.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "uncensored",
    "messages": [{"role": "user", "content": "Say hello in one short sentence."}],
    "max_tokens": 50
  }'

You should get a JSON response with a choices array. If you get a 401, the key is wrong or has a stray space. If you get a 404, the URL is wrong. You can also call GET https://api.deepseekapikey.com/v1/models with the same header to confirm the key is accepted.

Step 2: set up the proxy in Janitor AI

Open a chat with any character, then open the API or generation settings from the chat menu. Janitor AI offers its own built-in model and a few external options; pick the option for a proxy or custom OpenAI-compatible API.

1. Paste the full URL into the proxy URL field: https://api.deepseekapikey.com/v1/chat/completions. Janitor commonly expects the complete endpoint, not just the base. If the field hint in your version shows a URL that ends in a path, follow the hint.
2. Paste your key into the API key field. Do not add the word Bearer; Janitor adds the header itself.
3. Type uncensored as the model name. This is a free-text field, so spelling matters.
4. Save the settings, then reload the chat page. Many users skip the reload and wonder why the old settings are still active.
5. Send a short message. A reply from the character means the connection works.

If your Janitor version asks for a custom prompt or a jailbreak field, you do not need one for refusals on this API, because the model is tuned not to refuse lawful adult requests. Keep the character definition and the scenario clean and specific instead. Long, contradictory prompts waste tokens and confuse any model.

Tuning the generation settings

Janitor AI exposes a few sampling controls. These values are passed through to the model, so they have real effect:

• Temperature: 0.7 to 0.9 suits most roleplay. Higher values make replies more varied and less stable. If the character drifts or the text gets odd, lower it.
• Max new tokens: this caps the length of each reply. A value around 300 to 500 gives a paragraph or two. The model can use the rest of the 100,000-token window for output if you let it, but you rarely want that in chat.
• Context size: how much chat history Janitor sends each turn. The window is 100,000 tokens for prompt and reply combined, so anything up to that is possible. A bigger setting remembers more and costs more per message.

The model supports streaming, which makes replies appear word by word. If Janitor offers a streaming toggle, try it on. A streamed request holds its concurrency slot until it finishes, which only matters if you run several chats at once.

What a chat session costs

Billing is per token: $0.15 per 1M input tokens and $0.50 per 1M output tokens, from prepaid credit. Roleplay is input heavy, because every turn resends the character card and the conversation so far. Here is a worked example with a 12,000-token history and a 400-token reply:

• Input: 12,000 / 1,000,000 x $0.15 = $0.0018
• Output: 400 / 1,000,000 x $0.50 = $0.0002
• Per message: $0.0020, so 100 messages cost $0.20 and 1,000 messages cost $2.00.

The $0.50 trial credit covers about 250 messages at that size. If you let the history grow to 60,000 tokens, input alone is 60,000 / 1,000,000 x $0.15 = $0.009 per message, so the same 100 messages cost about $0.92. This is why a modest context setting is usually the better trade for ordinary chats.

Each response includes an X-Balance-USD header, and failed or refused requests are not charged. Credit never expires and there is no subscription. Top-ups run from $20 to $500 in any whole amount, with +5% from $50, +10% from $100 and +15% from $200. You can pay by crypto (USDT on TRC20, USDC on Base). The pricing page has the current details.

Fixing common errors

401 or "invalid key"

The key is wrong, was copied with a trailing space, or belongs to another service. A DeepSeek platform key will not work here, and this site's key will not work on DeepSeek's endpoint. Copy it again from your account page.

404 or "not found"

The URL is missing a part. Compare it character by character with https://api.deepseekapikey.com/v1/chat/completions. A doubled /v1/v1 or a missing /chat/completions are the usual causes.

429 with code concurrency or rate_limited

You sent too many requests at once or too many per minute. Trial keys allow 2 concurrent requests and 60 per minute; full keys allow 8 and 300. Do not click regenerate repeatedly while a reply is still streaming. A 429 with a Retry-After: 5 header means temporary overload, and waiting five seconds is enough.

Replies cut off or very short

Raise the max tokens setting. If replies stop mid-sentence, that is usually the cap in Janitor, not the API.

Model name errors

Use exactly uncensored, in lowercase. Names of other models will not map to anything here.

For the underlying parameters and examples in code, see the quickstart documentation. If you are still unsure about the difference between this service and DeepSeek's own, the article DeepSeek API: myths vs. facts goes through it.

Privacy and limits

Prompts sent through this API are not used for training. That is a statement about the API provider; Janitor AI has its own storage and policies for chats and characters, so read those separately.

The service is for lawful adult use, and there is one hard limit: no sexual content involving minors. Characters and scenarios that involve minors in that way will be refused, and no setting changes that. The API is text only, so image generation in Janitor will not come from this key.

More guides

• DeepSeek API Key in VS Code: Setup With Our API
• DeepSeek API Key in SillyTavern: Connect Our API
• DeepSeek API: Myths vs. Facts for Uncensored Access