How to Connect Janitor AI to DeepSeek (Official Model, 36% Off) — 2026 Proxy Setup with APIMaster

Why Janitor AI Needs a Proxy in the First Place

Janitor AI is a character-chat frontend, not a model provider. Its free tier runs on JLLM, an in-house model with no setup required, but every serious user eventually hits its limits — shorter context, generic writing, or just wanting the model they already know from elsewhere. To use Claude, GPT, DeepSeek or any third-party model, Janitor AI requires you to supply your own OpenAI-compatible endpoint and key through API Settings → Proxy. That's exactly the same mechanism whether you point it at the official OpenAI/Anthropic APIs directly, at OpenRouter, or at an aggregator like APIMaster — Janitor AI just forwards your chat as a standard chat/completions request to whatever Proxy URL you configured.

This is also why the search "janitor ai api key" or "janitor ai proxy" is really a provider-selection question in disguise — the Janitor AI side of the setup is four form fields; the actual decision is which backend you point them at.

Why APIMaster Is the Better Proxy Choice for Janitor AI

Most Janitor AI proxy guides point you at OpenRouter by default, simply because it's the most-mentioned option — not because it's the cheapest or safest one. Three things make APIMaster a better backend for the exact same setup:

• DeepSeek at 36% off, the model most Janitor AI users actually connect — OpenRouter's own usage ranking for Janitor AI shows DeepSeek variants taking most of the top 10 slots; APIMaster prices the official DeepSeek model at 36% off list, with OpenAI up to 90% off and Claude up to 85% off, at live per-model pricing on the marketplace — instead of paying OpenRouter's markup over official rates
• Fingerprint-verified models, not just a name on a label — in APIMaster's real-world checks, 43.8% of third-party API channels showed model mismatch; every channel we sell publishes its detection status on the model tester leaderboard, so "DeepSeek" in your proxy config is actually the official DeepSeek model, not a relabeled substitute
• One key for every model — DeepSeek, Claude, GPT, Kimi K3 and more all sit behind the same https://apimaster.ai/v1 endpoint, so trying a different model in Janitor AI is just editing the Model Name field, not signing up for a new provider
• Top up from $1 — pay as you go, no subscription, no minimum spend, with Alipay, WeChat Pay, USDT and credit card all supported
• Drop-in identical setup — if you've ever followed an OpenRouter proxy guide for Janitor AI, APIMaster uses the exact same four fields; only the Proxy URL and API Key change

How to Add APIMaster as a Proxy in Janitor AI

Step 1: Get an APIMaster API key

1. Register at https://apimaster.ai/register (or sign in at https://apimaster.ai/console/dashboard if you already have an account).
2. Open API Keys, click Create token, and copy it — you'll paste it into Janitor AI in a moment.
3. Optional but recommended: check the marketplace for the exact model id you want and its current per-token price before you commit credits.

Step 2: Open Janitor AI's proxy configuration

1. Log in to Janitor AI and open any chat.
2. Click the gear icon at the top of the chat, then API Settings.
3. Select Proxy, then click + Add Configuration.

Step 3: Fill in the four fields

Field	What to enter
Config Name	Anything you'll recognize later, e.g. APIMaster - DeepSeek
Model Name	The exact model id from the marketplace — e.g. deepseek-v4-pro, deepseek-v4-flash, claude-sonnet-4-6, gpt-5.5, kimi-k3 (copy-paste it, don't retype)
Proxy URL	https://apimaster.ai/v1/chat/completions (copy-paste it exactly — don't click it)
API Key	Paste your APIMaster key from Step 1, with no leading/trailing spaces

Leave Custom Prompt blank unless you already have a system prompt override you use intentionally.

Step 4: Save and test

1. Click Add, then Save Settings.
2. Fully refresh the page — Janitor AI's own guides call this step out specifically; a soft reload isn't always enough for a new proxy config to take effect.
3. Send a short test message, e.g. "hello", and confirm you get a normal reply.

If it replies correctly, your Janitor AI chat is now routed through APIMaster's https://apimaster.ai/v1 endpoint using the model you specified.

Which Model Name Should You Use?

Janitor AI's Model Name field must exactly match an id APIMaster actually serves — check the marketplace for the current list and pricing, since availability changes as new models launch. Commonly used ids include:

Model id	Family
deepseek-v4-pro / deepseek-v4-flash	DeepSeek — 36% off, most-used family for Janitor AI proxies
claude-sonnet-4-6 / claude-opus-4-7 / claude-opus-4-8	Claude (Anthropic)
gpt-5.4 / gpt-5.5	GPT (OpenAI)
kimi-k3	Kimi (Moonshot AI)
glm-5.2	GLM (Zhipu)
minimax-m3	MiniMax

To try a different model later, edit the existing configuration's Model Name field — you don't need a second Proxy URL or a new API key.

APIMaster vs OpenRouter for Janitor AI

	APIMaster	OpenRouter
Setup in Janitor AI	Same 4 fields (Config Name / Model Name / Proxy URL / API Key)	Same 4 fields
Pricing vs official list	DeepSeek 36% off, up to 90% off OpenAI, 85% off Claude	Small per-request markup over provider list price
Minimum top-up	$1	Varies by payment method, typically higher
Payment methods	Alipay, WeChat Pay, USDT, credit card	Credit card, crypto
Model authenticity check	Public fingerprint detection per channel	Not published

Neither is "wrong" — if you already have OpenRouter credits, keep using them. If you're deciding where to buy credits for Janitor AI specifically, the price and verification gap above is the practical difference.

Troubleshooting

"Unauthorized" error after saving the proxy This is almost always a Janitor AI session issue, not your APIMaster key — Janitor AI's own error docs attribute it to cookies not refreshing, browser bugs, or IP-based blocks. Log out and back in, try an incognito window, or switch browsers before assuming the key itself is bad. If it persists, verify the key independently with the API connectivity tester.

"Rate Limit / Network Error" Janitor AI's help center notes this usually means the upstream model provider is temporarily overloaded, not a problem with the proxy connection itself. Check the model's detection/status history on APIMaster, try again shortly, or temporarily switch to a different model id in the same configuration.

Model doesn't respond, or Janitor AI says the model isn't found The Model Name field must match an APIMaster marketplace id exactly — check for typos, extra spaces, or a model that's since been renamed on the marketplace.

Proxy URL doesn't seem to work It must be https://apimaster.ai/v1/chat/completions, including the /chat/completions suffix — pasting only https://apimaster.ai/v1 (without the suffix) is the most common mistake when copying from a different provider's shorter-looking URL.

FAQ

Is APIMaster free to use with Janitor AI? No — APIMaster is pay-as-you-go, not free. You top up from $1 and pay per token, but there's no subscription and no minimum spend, so cost stays proportional to how much you actually chat.

What's the exact Proxy URL for Janitor AI? https://apimaster.ai/v1/chat/completions. Copy it exactly — don't add or remove characters, and don't click the link inside Janitor AI's field.

Can I use the same setup for multiple models? Yes. Create one configuration and change the Model Name field whenever you want a different model — you don't need a separate Proxy URL or API key per model.

Does this replace Janitor AI's free JLLM option? No, JLLM still works with zero setup. This proxy setup is only needed if you specifically want Claude, GPT, DeepSeek, Kimi or another third-party model instead.

Is my chat data stored or used for training by APIMaster? No — API requests are forwarded to the underlying model provider and are not stored or used for training on APIMaster's side.

How do I know the model I'm chatting with is the real one? Model names aren't proof by themselves — a channel can silently serve a different model than advertised. Run a check on the free model tester before trusting a channel with real usage, the same way you'd check any third-party API channel.

Related guides: DeepSeek, Kimi K3, MiniMax M3 & GLM-5.2: 40% Off