Janitor AI in Russia: how to set up proxy, API, JLLM, and DNS — a complete guide for those who are having issues.
In Brief: Janitor AI is blocked in Russia at the DNS and IP level. To gain access, you need to set up a proxy or VPN, obtain an OpenAI API key, or use the alternative JLLM, and then enter the key in the character settings. The complete setup takes 15–20 minutes.
This guide is about technical methods to bypass blocks for a specific service. If you're looking for a ready-made solution without VPNs and API keys, check out alternatives with a Russian interface.
Janitor AI stopped opening for most users in Russia since spring 2023. The problem is not with your browser or provider — the service is indeed unavailable due to blocks at the DNS and routing level. However, access can be restored in a few steps: set up a bypass for the blocks, obtain an API key for the language model, and correctly link everything in the interface. Below is a step-by-step guide with specific commands and addresses.
Why Janitor AI Doesn't Work in Russia: Three Technical Reasons
The first reason is DNS blocking. Russian providers received an order not to resolve the domain janitorai.com. When you enter the address in your browser, the request goes to the provider's DNS server, which returns an empty response or redirects to a placeholder. You can check this with the command nslookup janitorai.com in the Windows command line or macOS terminal — if the response does not contain an IP address, the problem is indeed with DNS.
The second reason is IP blocking at the routing level. Even if you set an alternative DNS (for example, 1.1.1.3 from Cloudflare), packets to Janitor AI servers may drop on border routers. Roskomnadzor has blacklisted the IP address ranges of the hosting, and the provider physically does not allow the traffic through.
The third reason concerns the API. Even if the site opens via VPN, the chat requires an OpenAI API key or an alternative model. OpenAI stopped registering new accounts from Russia in March 2023 and blocks Russian cards. Old keys work, but obtaining a new one legally is not possible. An alternative is to use JLLM (the built-in free model of Janitor AI) or third-party API providers that support Russian cards.
Step 1: Setting Up DNS and Initial Access Check
Start by changing the DNS servers. Open the network connection settings: in Windows — "Settings" → "Network & Internet" → "Change adapter settings," right-click on the active connection → "Properties" → "Internet Protocol Version 4." In macOS — "System Preferences" → "Network" → select the connection → "Advanced" → DNS tab.
Enter the following DNS servers:
- 1.1.1.3 and 1.0.0.3 — Cloudflare for Families (blocks malicious sites but does not censor based on Roskomnadzor lists)
- 8.8.8.8 and 8.8.4.4 — Google Public DNS (works more reliably, but Google complies with some government requests)
- 94.140.14.14 and 94.140.15.15 — AdGuard DNS (servers in the Netherlands, good speed for the European part of Russia)
After changing the DNS, open the command line and execute ipconfig /flushdns (Windows) or sudo dscacheutil -flushcache (macOS) to clear the cache. Try to open janitorai.com. If the page loads — DNS blocking has been bypassed. If you see the error "ERR_CONNECTION_TIMED_OUT" or "Site unavailable," proceed to the next step.
Step 2: Choosing and Setting Up a VPN or Proxy Server
If changing the DNS didn't help, you need a VPN or proxy to mask your real IP. Free VPNs (Proton VPN, Windscribe) work, but often throttle speeds to 1–2 Mbps, making communication with the chatbot painfully slow. Paid services are more stable: Mullvad VPN (5 euros/month, payment in cryptocurrency without registration), Surfshark (supports Russian cards through intermediaries), IVPN.
For Janitor AI, an HTTP/HTTPS proxy is sufficient. Find a public proxy server on sites like free-proxy-list.net (filter by EU/US countries, HTTPS protocol, uptime >90%). Enter the proxy in your browser settings: in Chrome/Edge — "Settings" → "System" → "Open your computer's proxy settings" → "Network setup" → enter the address and port (for example, 185.162.235.94:3128). In Firefox — "Settings" → "General" → "Network Settings" → "Settings" → "Manual proxy configuration."
Check if it works: open whoer.net and ensure that your IP address has changed, and the country is identified as not Russia. Then open janitorai.com. If the site loads, the block has been bypassed. Register or log in via Google/Discord.
Step 3: Obtaining an OpenAI API Key or Setting Up JLLM
Janitor AI does not store its own language model — it uses APIs from third-party services. You have three options:
Option A: OpenAI API (if you already have an account registered before March 2023). Log in to platform.openai.com via VPN. Go to the "API keys" section → "Create new secret key." Copy the key in the format sk-proj-aBcD1234... — it is shown only once. The minimum balance for operation is $5, and top-ups are only possible with foreign cards or through intermediaries (Telegram bots like @openai_pay_bot, commission 15–25%).
Option B: JLLM (the free model of Janitor AI). In the character settings, choose "JLLM" instead of OpenAI. The model is weaker than GPT-3.5: it often loses context after 10–15 messages, may repeat itself, and struggles with complex instructions. However, it does not require an API key and works immediately. Suitable for simple role-playing scenarios.
Option C: Alternative API Providers. Services like openrouter.ai, poe.com (API mode), together.ai provide access to models Claude, Llama, Mistral. Registration is easier, and some accept cryptocurrency. However, Janitor AI officially supports only OpenAI API and JLLM — for other providers, you need workarounds (proxy scripts in Python that emulate OpenAI request formats).
Step 4: Linking the API Key to the Character in Janitor AI
Open the character catalog on janitorai.com (if you need specific genres, use filters or check collections of romantic characters). Choose a character, click "Chat." If the API is not yet set up, you will see a window "API Setup Required."
Click "Setup API" → choose "OpenAI" or "JLLM." If you selected OpenAI, paste the copied key into the "API Key" field. Choose the model gpt-3.5-turbo (cheaper, $0.002 per 1000 tokens) or gpt-4 (smarter, but $0.03 per 1000 tokens — a $5 balance will run out after 50–100 messages). Click "Save."
If you chose JLLM, simply click "Confirm" — no additional settings are needed. The model will connect automatically.
Send a test message. If the character responds — everything works. If you see the error "Invalid API key" — check if the key has expired on platform.openai.com (in the "Usage" section). If you see the error "Rate limit exceeded" — the request limit has been exceeded (the free tier of OpenAI is limited to 3 requests per minute), wait 60 seconds or top up your balance to increase the limit.
Step 5: Optimizing Settings for Stable Operation
To prevent the chat from dropping and the model from "forgetting" context, adjust the parameters in the character interface. Click on the gear icon in the upper right corner of the chat.
Context Length — how many previous messages the model sees. For gpt-3.5-turbo set it to 4096 tokens (about 50–60 messages), for gpt-4 you can set it to 8192. For JLLM, no more than 2048 — otherwise, the model starts to "drift."
Temperature — creativity of responses. 0.7–0.9 for role-playing games (varied reactions), 0.3–0.5 for informational queries (more predictable responses).
Frequency Penalty — penalty for repeating phrases. Set it to 0.3–0.5 to prevent the character from getting stuck on the same expressions.
Save the settings. If you're using a VPN, enable the "Always-on" mode in the VPN app to prevent the connection from dropping when switching between tabs.
Comparison of Access Methods to Janitor AI from Russia
| Method | Cost | Speed | Setup Complexity | Stability |
|---|---|---|---|---|
| Changing DNS | Free | No loss | Low (5 minutes) | Works for 20–30% of users |
| Free VPN | Free | 1–3 Mbps | Low (app installation) | Average (drops under load) |
| Paid VPN | 300–800₽/month | 20–50 Mbps | Low | High |
| HTTP Proxy | Free or 100–300₽/month | 5–15 Mbps | Medium (manual setup) | Average (public proxies are often blocked) |
| OpenAI API | From $5 (~500₽) + commission | No effect | Medium (foreign account needed) | High |
| JLLM (built-in model) | Free | No effect | Low | Average (lower quality responses) |
Common Errors During Setup and How to Fix Them
Error 1: "API key is invalid or expired." Reason — the key was copied incompletely (truncated space at the end) or the OpenAI account is blocked for violating usage rules. Solution: generate a new key on platform.openai.com, ensuring you copy it entirely (it starts with sk-). If the account is blocked, you will need to register a new one via VPN with an email from another country.
Error 2: the character responds off-topic or repeats the same thing. Reason — too low Context Length (the model does not see previous messages) or high Frequency Penalty. Solution: increase Context Length to 4096, lower Frequency Penalty to 0.3, restart the chat (click "New Chat").
Error 3: the site opens, but the chat does not load — the loading icon keeps spinning. Reason — WebSocket connections are blocked by the provider or proxy. Solution: use a VPN that supports all protocols (not just HTTP/HTTPS), disable browser extensions that may block scripts (uBlock Origin, Privacy Badger).
Error 4: "Rate limit exceeded — too many requests." Reason — the API request limit has been exceeded (for the free OpenAI account — 3 requests per minute, for paid tier 1 — 3500 per minute). Solution: wait 60 seconds between messages or top up your OpenAI balance by at least $10 to switch to tier 2 with a limit of 10,000 requests per minute.
Error 5: VPN works, but Janitor AI still won't open. Reason — DNS requests are leaking outside the VPN (DNS leak). Solution: check at dnsleaktest.com — if you see DNS servers from Russian providers, enable the "DNS leak protection" option in the VPN settings or manually enter DNS (1.1.1.1, 8.8.8.8) in the network adapter settings.
Alternatives to Janitor AI for Users from Russia
If setting up Janitor AI seems too complicated or you don't want to spend on VPNs and APIs, there are services with similar functionality available without blocks. For example, platforms with ready-made AI characters that work directly from Russia, do not require API keys, and offer free plans for exploration.
Another option is local models that run on your computer (LM Studio, Oobabooga Text Generation WebUI). They require powerful hardware (a graphics card with 8+ GB VRAM for models like Llama 2 13B), but are completely autonomous — no blocks, APIs, or subscriptions. Look for installation instructions in the GitHub repositories of the projects.
A third path is Telegram bots based on GPT-3.5 or Claude (for example, @ChatGPT_RU_bot). They work through developers' proxies, the API is already set up, and you only pay for messages (1–3₽ per request). The downside is that there is no customization of characters, just a standard assistant.
Legal and Ethical Aspects of Using VPN in Russia
Using a VPN for personal purposes in Russia is legal — the law only prohibits providing VPN services without registration with Roskomnadzor and integration with the FGIS (blocking system). Individuals do not face fines for connecting to foreign VPNs. However, it is important to be aware of the risks: some VPN services log traffic and may share data upon request from authorities.
If you work with confidential information, choose a VPN with a "no-logs" policy (Mullvad, IVPN, ProtonVPN) and payment in cryptocurrency without personal identification. Avoid free VPNs from China and Russia — they often embed ads, sell data, or contain malicious code.
The ethical aspect: AI chats like Janitor AI use models trained on data from the internet, including works by authors without their consent. If you create content using AI, make it clear and do not present it as entirely original text. Respect community rules — many platforms prohibit publishing AI-generated images or texts without a label.
Frequently Asked Questions
Can I use Janitor AI for free without an API key?
Yes, if you choose the built-in JLLM model in the character settings. It is free, does not require registration with OpenAI, and works immediately after logging into the site. The quality of responses is lower than that of GPT-3.5 — the model struggles to maintain context in long dialogues and may generate repetitive phrases. For short role-playing scenarios or getting acquainted with the platform, JLLM is suitable, but for serious projects, it's better to use the paid API.
Which VPN works best with Janitor AI in Russia?
According to user reviews, the most stable are Mullvad VPN (servers in Sweden and Germany, speed 30–50 Mbps, payment in cryptocurrency), Surfshark (supports the WireGuard protocol, which is harder to detect), and ProtonVPN (free plan with speed limits but no traffic limits). Avoid VPNs with servers only in the USA — they have higher latency (ping 150–200 ms), making communication with the chat less comfortable.
Why does Janitor AI consume the OpenAI balance in just a few days?
Each message in the chat is an API request charged by the number of tokens (approximately 1 token = 0.75 words). If the Context Length is set to 8192 tokens, the model processes the entire dialogue history each time, even if you send a short message. Solution: lower the Context Length to 2048–4096, use gpt-3.5-turbo instead of gpt-4 (15 times cheaper), delete old chat threads using the "Delete Chat" button. With moderate use (20–30 messages