How to Use DeepSeek on Janitor AI: Complete Setup Guide for 2026

Janitor AI is a popular platform for chatting with AI characters, but its default model has limits. Connecting DeepSeek's API gives you access to a more capable model for free or at very low cost. Here's how to set it up.

What You Need Before Starting

Three things:

A Janitor AI account (sign up at janitorai.com)
A DeepSeek API key (free to create, pay-per-use)
A few minutes to configure the connection

The DeepSeek API is paid, but the costs are tiny for casual use. Typical character chats run a few cents per session. DeepSeek gives new API users free credits, so you can test the setup without spending anything.

Step 1: Get Your DeepSeek API Key

Go to platform.deepseek.com and sign up. You'll need an email address or phone number. After verifying your account:

1. Click "API Keys" in the left sidebar.

2. Click "Create new API key."

3. Give it a name like "Janitor AI."

4. Copy the key immediately. It starts with "sk-" and you won't be able to see it again after closing the window.

Keep this key private. Anyone with your API key can use your DeepSeek account and you'll be billed for their usage.

Step 2: Add Funds to Your Account

Go to the "Billing" section in the DeepSeek platform. Add a small amount, $5 is plenty to start. DeepSeek V4-Flash costs $0.28 per million input tokens and $1.10 per million output tokens. A character chat session with a few dozen messages typically uses 10,000-50,000 tokens total, costing well under a dime.

If you received free credits when signing up, you can skip this step and add funds later when the credits run out.

Step 3: Configure Janitor AI

Log into Janitor AI and go to the API settings:

1. Click the menu icon (three lines) in the top left.

2. Select "API Settings" from the menu.

3. Under "API Provider," choose "Custom" or "OpenAI Compatible."

4. Set the API endpoint to: https://api.deepseek.com/v1/chat/completions

5. Paste your DeepSeek API key into the API key field.

6. For the model name, enter: deepseek-v4-flash (for fast, cheap responses) or deepseek-v4-pro (for better quality, higher cost).

Some Janitor AI interfaces show these fields under different names. If you see "Reverse Proxy" instead of "Custom API," use the same endpoint. The key settings are the endpoint URL and the model name.

Step 4: Choose the Right Model

Your model choice affects both the chat quality and your costs:

deepseek-v4-flash (recommended for Janitor AI):

Fast responses, low latency
Good for character roleplay and casual chat
Very cheap: roughly $0.28/million input tokens
The default choice for most users

deepseek-v4-pro:

Better at complex reasoning and longer context
More natural conversation flow
Higher cost but still cheaper than OpenAI models
Worth it if you're having in-depth, multi-turn conversations

For most character chats on Janitor AI, Flash is perfectly adequate. Characters with complex backstories or long-running narratives might benefit from Pro.

Step 5: Test the Connection

After saving your API settings, start a chat with any character. Send a simple message like "Hi, how are you?" and check:

The response comes back within a few seconds.
The character's personality comes through in the reply.
There are no error messages about API keys or billing.

If you see an error about the API key, double-check that you pasted the full key starting with "sk-". If you see a billing error, add funds to your DeepSeek account.

Step 6: Adjust Settings for Better Chats

In Janitor AI's generation settings, tweak these for DeepSeek:

Temperature: 0.7 to 0.9. Lower values make responses more predictable. Higher values add variety. For character chat, 0.8 is a good starting point.
Max tokens: 500 to 1000. This controls response length. Too low and characters give one-word answers. Too high and you'll burn through your API credits faster.
Top P: Leave at 0.9 or the default. This setting doesn't need much adjustment for character chat.

Troubleshooting

"Invalid API key" error: Your key may have been copied incorrectly. Generate a new key and try again. Make sure there are no spaces before or after the key.

Slow responses: If DeepSeek's servers are busy, switch from deepseek-v4-pro to deepseek-v4-flash. Flash has more capacity and responds faster during peak hours.

Characters forgetting context: DeepSeek's context window on Janitor AI is limited by the platform, not the model. If conversations start losing context after many messages, start a new chat. The model itself supports up to 1 million tokens, but Janitor AI may cap it lower.

Repetitive responses: Lower the temperature to 0.6. If the character keeps repeating phrases, they're stuck in a loop. Start a new chat or edit your last message to break the pattern.

Cost Expectations

DeepSeek is one of the cheapest APIs for Janitor AI. For a typical user chatting a few hours per week:

Flash model: $1-$3 per month
Pro model: $3-$8 per month
Heavy daily use: $10-$20 per month

Compare that to OpenAI's API on Janitor AI, which can easily run $30-$80 per month for similar usage. DeepSeek's pricing is the main reason people switch.

DeepSeek - AI Assistant

DeepSeek

DeepSeek is an AI assistant app that offers free access to its latest reasoning model.

Productivity AI Chatbot

Download (26.6 MB)

The DeepSeek app itself is free and separate from the API. You only pay for API usage through Janitor AI or other third-party tools, not for the official app.