How to Set Up DeepSeek on Janitor AI (Complete 2026 Guide)

You want to set up DeepSeek on Janitor AI and actually get it working not just read a guide that skips the hard parts, you’re in the right place. This is the most complete tutorial available, covering direct DeepSeek API setup, free proxy options through Nebula Block (MegaNova), troubleshooting every common error, and the exact settings that produce the best responses. By the end, your setup will be fully running with optimized generation parameters.

What You Need to Know First: How Janitor AI Actually Works

Before touching any settings, understand this one fact that most guides skip:

Janitor AI does not host or run any AI model. It is purely a chat interface, a front end  that routes your messages to whichever external API you configure. When you “use DeepSeek on Janitor AI,”

 what’s actually happening is:

• You type a message in Janitor AI.
• Janitor AI sends that message to DeepSeek’s API servers.
• DeepSeek processes it and returns a response.
• Janitor AI displays that response in the chat.

This is why an API key is required. Without it, Janitor AI has nowhere to send your messages. This is also why most setup failures happen with incorrect API endpoints, expired keys, or missing proxy configuration.

Two Ways to Connect DeepSeek to Janitor AI

There are two main methods, depending on whether you want to pay directly or use a free proxy:

Method	Cost	Best For
Direct DeepSeek API	Pay-per-token (very affordable)	High quality, reliable, production use
Nebula Block (MegaNova) Free Proxy	Free tier available	Casual use, testing, beginners

Both methods are covered below. Start with whichever fits your situation.

First Method : Direct DeepSeek API Setup

Create Your DeepSeek Account

• Go to platform.deepseek.com
• Click Sign Up and register with your email address
• Verify your email through the confirmation link
• Log in to your DeepSeek dashboard
  

 Add Credits to Your Account

DeepSeek uses a pay-per-token billing model. Before generating an API key, add a small amount of credits (even $1–$5 is enough to get started):

• In your dashboard, go to Billing or Top Up
• Add credits using a supported payment method
• Wait for the balance to reflect (usually instant)

Why add credits first? 

An API key without balance will authenticate but return errors mid-conversation. Adding credits upfront avoids this frustrating issue.

Generate Your DeepSeek API Key

• In your DeepSeek dashboard, navigate to API Keys (sometimes listed under “Developer” or “Account”)
• Click Create New API Key
• Give it a name (e.g., “JanitorAI”)
• Copy the key immediately it starts with sk and will only be shown once
• Store it somewhere secure (a password manager works well)

Security warning:

 Never post your API key publicly  in screenshots, forums, Reddit threads, or anywhere else. Anyone with your key can use your credits.

Configure DeepSeek Inside Janitor AI

• Log in to your Janitor AI account
• Click your profile icon in the top right
• Go to Settings → API Settings
• Under the API type selector, choose Custom (not OpenAI or Claude)
• Fill in the following fields:
  • API Key: Paste your DeepSeek API key (sk-…)
  • API Endpoint / Proxy URL: https://api.deepseek.com/v1/chat/completions
  • Model: deepseek-chat (for DeepSeek-V3, recommended) or deepseek-reasoner (for DeepSeek-R1)
• Click Save
• Click Test Connection to verify everything works

If the test returns a green confirmation, you’re set. If it fails, see the troubleshooting section below.

Choose Your DeepSeek Model

DeepSeek currently offers two main models accessible via API:

Model Name	API String	Best For
DeepSeek-V3	deepseek-chat	General conversation, roleplay, creative writing
DeepSeek-R1	deepseek-reasoner	Complex reasoning, detailed responses

For most Janitor AI users, DeepSeek-V3 (deepseek-chat) is the better choice. It’s faster, more conversational, and handles roleplay scenarios naturally. DeepSeek-R1 is better if you want more analytical, step-by-step reasoning in responses.

Second Method : Free DeepSeek Proxy via Nebula Block (MegaNova)

If you don’t want to add billing details to DeepSeek directly, you can use Nebula Block (now operating under the MegaNova brand), which offers free access to DeepSeek models through an OpenAi compatible endpoint.

Note: Free tier models are lighter, quantized versions. They work well for casual roleplay but may produce shorter or slightly less detailed responses than the paid direct API.

Create a Nebula Block / MegaNova Account

• Go to console.meganova.ai
• Sign up  no credit card required for the free tier
• Verify your email and log in

 Generate Your Nebula Block API Key

• In the dashboard, navigate to API Keys
• Click Generate New Key
• Copy the key (it starts with sk-)

 Find Your Model Name and Endpoint

• Go to the Serverless Models page
• Find DeepSeek-V3-0324 (the free version is labeled DeepSeek-V3-0324-Free)
• Switch to the API tab and select cURL to see your endpoint details
• Note down:
  • Model name: deepseek-ai/DeepSeek-V3-0324-Free
  • API Endpoint: https://inference.meganova.ai/v1/chat/completions

Configure Janitor AI with the Proxy

• In Janitor AI, go to Settings → API Settings
• Select Custom as your API type
• Fill in:
  • API Key: Your MegaNova key
  • Proxy URL / Other API URL: https://inference.meganova.ai/v1/chat/completions
  • Model: deepseek-ai/DeepSeek-V3-0324-Free (type this exactly)
• Save your settings

Enable Proxy on Your Characters

This is a step many guides miss entirely. Even after configuring the proxy in settings, individual characters may have proxy usage disabled.

If you’re getting no responses from a character:

• Open the character you want to chat with
• Look for a settings or gear icon on the character page
• Find the Allow Proxies toggle and make sure it is enabled
• Return to your chat and try again

GeneBest ration Settings for DeepSeek on Janitor AI

Once your API is connected, your generation settings determine the quality and style of responses. These are the recommended values:

Setting	Recommended Value	What It Does
Temperature	0.7 – 0.85	Controls creativity. Higher = more varied, lower = more predictable
Max Tokens	1000 – 2000	Maximum length of each response
Top-P	0.9	Nucleus sampling  keeps outputs coherent while allowing variety
Presence Penalty	0.4 – 0.6	Reduces repetition of topics already mentioned
Frequency Penalty	0.3 – 0.5	Reduces word-level repetition within a response

Quick Setting Profiles

For immersive, creative roleplay:

• Temperature: 0.85
• Max Tokens: 1500
• Top-P: 0.92
• Presence Penalty: 0.5

For grounded, consistent dialogue:

• Temperature: 0.70
• Max Tokens: 1000
• Top-P: 0.88
• Presence Penalty: 0.3

For long-form, detailed storytelling:

• Temperature: 0.80
• Max Tokens: 2000
• Top-P: 0.90
• Presence Penalty: 0.6

Pro tip: If responses feel repetitive, increase Presence Penalty to 0.7. If they feel random or incoherent, lower the temperature to 0.65.

Troubleshooting: Common Errors and Exact Fixes

Error: “Invalid API Key” or “Authentication Failed”

Cause: The API key was entered incorrectly, has expired, or was regenerated.

Fix:

• Go back to your DeepSeek (or MegaNova) dashboard
• Regenerate a new API key
• Delete the old key from Janitor AI settings completely
• Paste the new key and save

Error: “Connection Failed” or “Unable to reach API”

Cause: Wrong endpoint URL, or the API service is temporarily down.

Fix:

• Double-check your endpoint URL for typos:
  • Direct DeepSeek: https://api.deepseek.com/v1/chat/completions
  • MegaNova proxy: https://inference.meganova.ai/v1/chat/completions
• Check status.deepseek.com for any ongoing outages
• Try again after a few minutes

Error: “Insufficient Balance” or Responses Stop Mid-Conversation

Cause: Your DeepSeek account has run out of credits.

Fix:

• Log in to your DeepSeek dashboard
• Navigate to Billing and top up your balance
• Return to Janitor AI  the next message will work normally

Error: No Response from a Specific Character

Cause: The character has proxy usage disabled (very common with MegaNova setup).

Fix:

• Open the character’s page on Janitor AI
• Go to character settings
• Enable the Allow Proxies toggle
• Try chatting again

Error: “Rate Limit Exceeded”

Cause: Too many requests sent in a short period, especially on free tiers.

Fix:

• Wait 30–60 seconds before sending the next message
• If this happens frequently, consider upgrading to a paid DeepSeek API tier
• Reduce Max Tokens in your settings to use fewer resources per message

Error: “Model Not Found”

Cause: Incorrect model string in your settings.

Fix: Use exactly one of these strings:

• deepseek-chat (for direct DeepSeek API)
• deepseek-reasoner (for DeepSeek R1 via direct API)
• deepseek-ai/DeepSeek-V3-0324-Free (for MegaNova free tier)

Do You Need a Proxy?

A common point of confusion is when you actually need a proxy. Here’s a clear breakdown:

You likely do NOT need a proxy if:

• You’re connecting directly to DeepSeek’s own API
• You’re in a region where DeepSeek is accessible (most of the world)
• Your connection test in Janitor AI succeeds without one

You may need a proxy if:

• DeepSeek’s API is restricted or blocked in your country
• You want to use a free alternative like MegaNova instead of paying DeepSeek directly
• You’re getting consistent connection failures despite a valid API key

The MegaNova method described above functions as a proxy that routes your requests through their infrastructure to the DeepSeek model, which also means it works even in regions where direct DeepSeek access is limited.

DeepSeek vs Other APIs on Janitor AI

Here’s how DeepSeek compares to the other popular API choices for Janitor AI:

API	Cost	Reasoning Quality	Roleplay Quality	Free Option
DeepSeek V3	Very low	Excellent	Very good	Via MegaNova
DeepSeek R1	Low	Outstanding	Good	Via MegaNova
OpenAI GPT-4o	Higher	Excellent	Very good	No
Claude (Anthropic)	Medium	Excellent	Excellent	No

DeepSeek’s main advantages for Janitor AI users are its cost-effectiveness and the availability of a free proxy option through MegaNova. For users who want the most immersive roleplay at the lowest cost, DeepSeek V3 is currently the strongest choice.

Quickstart Checklist

Use this checklist to verify your setup is complete:

• Created a DeepSeek account (or MegaNova account for free tier)
• Generated an API key starting with sk
• Added credits to DeepSeek (if using direct API)
• Opened Janitor AI Settings → API Settings
• Selected Custom API type
• Pasted API key correctly
• Entered the correct endpoint URL
• Entered the correct model name
• Saved settings
• Ran connection test successfully
• Enabled “Allow Proxies” on desired characters
• Adjusted generation settings (Temperature, Max Tokens, etc.)

Once every box is checked, your DeepSeek setup on Janitor AI is complete and optimized. You should now have fast, high-quality AI responses in all your conversations.

FAQS

Is DeepSeek free to use on Janitor AI?

The direct DeepSeek API requires payment per token, but the cost is very low often less than $0.01 for a full conversation. Alternatively, Nebula Block (MegaNova) offers a free tier using DeepSeek models that requires no payment at all, making it the best option for users who want to try everything before spending any money.

Which DeepSeek model should I use for Janitor AI?

For most users, DeepSeek-V3 (deepseek-chat) is the best choice. It’s fast, conversational, and handles roleplay naturally. DeepSeek-R1 is better for users who want very detailed analytical responses, but it can sometimes feel less natural in casual conversation.

Why is my DeepSeek working but responses are short or cut off?

Your Max Token’s setting is too low. Increase it to at least 1000, and ideally 1500–2000 for longer, more immersive responses.

Can I use DeepSeek on Janitor AI on mobile?

Yes. The Janitor AI website is mobile-accessible. The API setup process is identical on mobile go to Settings → API Settings in your browser and follow the same steps.

Why does my setup work for some characters but not others?

Individual characters can have the Allow Proxies option disabled. You need to enable this setting on each character where you want to use your custom API.

Is it safe to enter my DeepSeek API key on Janitor AI?

Janitor AI is a well-established platform used by millions of users. However, as a general security practice: never use the same API key across multiple platforms, monitor your DeepSeek usage dashboard for any unexpected activity, and regenerate your key if you ever suspect it’s been compromised.