Hello, Janitorai kept flagging the original tutorial I made; check out "Important Bullshit - Click on me". Click on the arrow to go back to Janitorai.
Last updated: August 16th 2026
Janitorai kept flagging my tutorial because it said "Content moderation failed: The text promotes external proxy services and directs users to create API keys and add funds, which is a suspicious external service solicitation rather than ordinary story content." so damn, excuse a bitch for trying to teach people how to use proxies because when I was learning proxies, I couldn't find a tutorial that was current and accurate.So enjoy this I guess.
Color Key:
Suggested Models
Name Of Model
PSA Alert
Proxy URL - not clickable
1) Head over to platform.claude.com and make an account.2) On the left there will be an area called "API Keys" Click it, then "Create Key" which will be at the top right.3) Make a name for the key. Set "Expires" as never and then click "add"4) After creating it, it'll give you the key right away. You can not view this key again.5) Before it'll work, click on "claude console" at the top left and click on "Add funds" button in the "organization credits" box.6) Once all steps are completed, go into a chat on Janitorai and click on the purple button that says "using janitor", and under "Provider" there are two boxes: one that says "Model" and one that says "API Key". Preset is lovely, you just click on "preset" and then "select model" and it will automatically show you the models you can choose and then you click on the model you want. Or if you prefer to type in the model, click on "Custom" and choose one of the suggested models below and copy the entire thing and put it in the box under "custom". Put the API Key where it says API key.Suggested models: claude-sonnet-5 | claude-haiku-4-5-20251001sonnet is cost-efficient but a good middle ground for more expressive bots.
haiku is the cheapest, but can lose some personality/nuance in longer or more detailed bots.
1) Head over to platform.claude.com and make an account.2) On the left there will be an area called "API Keys" Click it, then "Create Key" which will be at the top right.3) Make a name for the key. Set "Expires" as never and then click "add"4) After creating it, it'll give you the key right away. You can not view this key again.5) Before it'll work, click on "claude console" at the top left and click on "Add funds" button in the "organization credits" box.6) Once all steps are completed, go into a chat on Janitorai and click on the purple button that says "using janitor" and, under provider, click on "Proxy". Click on the button that says " + New". There, you put the name of the proxy, one of the suggested models below, and lorebarys "proxy url" which for Claude is "https://api.lorebary.com/anthropic" and the API key that you copied earlier. Custom prompts: head over here, and you will see prompts that you can copy and place in "custom prompt". There are global prompts and custom prompts; I explain the difference at the bottom in "Things to keep in mind."Suggested models: claude-sonnet-5 | claude-haiku-4-5-20251001sonnet is cost-efficient but a good middle ground for more expressive bots.
haiku is the cheapest, but can lose some personality/nuance in longer or more detailed bots.
PSA: DeepSeek API service is expected to adopt a peak-valley pricing strategy starting in mid-July, with peak-hour prices being twice the regular price, applicable to all billing items.
Peak hours (in UTC): 1:00–4:00 AM and 6:00–10:00 AM.
(UTC+8 equivalent: 9:00 AM–12:00 noon and 2:00–6:00 PM.)PSA 2: Effective August 16 Deepseek V4 API has new pricing, both flash and pro. View the update here at the end of the page.1) Head over to platform.deepseek.com and make an account.2) On the left, there will be an area where you can click "API Keys" and after you click that, click "create new API key".3) It will ask for a name for the key, I used something simple like: Heisenberg. And then click "create api key"4) after creating it will give you the new API key right away, BUT YOU CAN NOT VIEW THIS KEY AGAIN, SO COPY IT AND SAVE IT SOMEWHERE SAFE.5) After its created, just below "API Keys" click on "Top Up" and follow the steps there.6) Once all steps are completed, go into a chat on Janitorai and click on the purple button that says "using janitor" and, under provider, click on "Proxy". Click on the button that says " + New". There, you put the name of the proxy, one of the suggested models below, and lorebarys proxy url which for Deepseek is https://api.lorebary.com/deepseek and the API key that you copied earlier. Custom prompts: head over here, and you will see prompts that you can copy and place in "custom prompt". There are global prompts and custom prompts; I explain the difference at the bottom in "Things to keep in mind."Suggested models: deepseek-v4-flash | deepseek-v4-pro
deepseek-v4-flash is faster, cheaper, good for casual/expressive RP.
deepseek-v4-pro is slower, pricier, stronger reasoning/consistency, better for detail-heavy or long-context bots.
PSA: As of April 2026, Google restructured their free tier - Flash models are still free (rate-limited), but Pro models now require paid billing set up. No more free Pro access. According to Gemini.1) Head over to aistudio.google.com sign in or create a new email.2) In the top left corner, there is a button that says "create API Key".3) Name your key whatever you want, simple is better.4) In the drop-down box below "name your key" that says "Choose an imported project" click "create project." Name it whatever you named your key. Your new API Key will pop up under "API Keys." Click the key which will start with "...." and when you click it, it prints up the full UI that allows you to copy that API key and paste it into the API configuration on janitorai.5) If you want to use gemini-3.5-flash, you're done it still works free with rate limits. If you want gemini-2.5-pro, you'll need to enable billing on your project first which I can not help with since I have never done this myself. (Google no longer allows free Pro access as of April 2026 according to Gemini).6) Once all steps are completed, go into a chat on Janitorai and click on the purple button that says "using janitor" and, under provider, click on "Proxy". Click on the button that says " + New". There, you put the name of the proxy, one of the suggested models below, and lorebarys proxy url, which for Gemini is https://api.lorebary.com/aistudio and the API key that you copied earlier. Custom prompts: head over here, and you will see prompts that you can copy and place in "custom prompt". There are global prompts and custom prompts; I explain the difference at the bottom in "Things to keep in mind."Suggested models: gemini-3.5-flash | gemini-2.5-pro
gemini-3.5-flash is fast, proactive, drives the scene forward, less nuance/subtext and it's the one that's actually still free.
gemini-2.5-pro is slower and pricier but picks up on subtext and emotional depth better, better for slow-burn writing now requires billing to be enabled.
1) Head over to openrouter.ai sign in or create a new email.2) After you sign in, in the top left corner, there will be your profile. Hover over your profile icon and click "workspaces", on the left, there will be a column that says "API Keys". Click on that and make a new key. Name the key whatever you want, leave the expiration as no expiration, and leave the key limit alone. Be sure to copy the API Key to your notepad because you can not view the key again once made.3) Near the top right, you will see a link that says "models" after you click on that on the middle of the screen there will be a box that says "All variants" when you click on that you will see "Free". Click that.4) Once you find an AI, all the names will have (free) after it. Click on the name; underneath the name will be something that looks like the provider and then the ai name. For example, for NVIDIA: Nemotron 3.5 Lightning (free) it will nvidia/nemotron-3.5-lightning:free underneath the name. Copy that.5) Once all steps are completed, go into a chat on Janitorai and click on the purple button that says "using janitor" and, under provider, click on "Proxy". Click on the button that says " + New". There, you put the name of the proxy and a model, and lorebarys proxy url, which for Open Router is https://api.lorebary.com/openrouter and the API key that you copied earlier. Custom prompts: head over here, and you will see prompts that you can copy and place in "custom prompt". There are global prompts and custom prompts; I explain the difference at the bottom in "Things to keep in mind."
1) Head over to openrouter.ai sign in or create a new email.2) After you sign in, in the top left corner, there will be your profile. Hover over it and click "credits".3) Add credits.4) After credits are added, hover over your profile icon and click "workspaces", on the left, there will be a column that says "API Keys". Click on that and make a new key. Name the key whatever you want, leave the expiration as no expiration, and leave the key limit alone. Be sure to copy the API Key to your notepad because you can not view the key again once made.5) Near the top right, you will see a link that says "models" after you click on that, on the left there will be a column of tabs, click on "categories," and underneath will be "roleplay" click on that.6) Once you compare the ai's good for roleplay, when you click the ai you want, underneath the name will be something that looks like the provider and then the ai name. For example, for Google: Gemini 3.7 Flash it will say google/gemini-3.7-flash underneath the name. Copy that.7) Once all steps are completed, go into a chat on Janitorai and click on the purple button that says "using janitor" and, under provider, click on "Proxy". Click on the button that says " + New". There, you put the name of the proxy and a model, and lorebarys proxy url, which for Open Router is https://api.lorebary.com/openrouter and the API key that you copied earlier. Custom prompts: head over here, and you will see prompts that you can copy and place in "custom prompt". There are global prompts and custom prompts; I explain the difference at the bottom in "Things to keep in mind."
According to a community manager and a lorbary dev that I talked to, jailbreak and foremarkdown sometimes conflicts with each other.According to a Lorebary dev and community manager I spoke with, using jailbreak on providers that offer more relaxed TOS (Llama-based models are one example) could cause the AI to act funky.
From my own experience: Claude and Gemini can handle subtle sexual scenes or hints, but anything full-on sexual will likely violate TOS.This is all researched by me. I have tried out a lot of different proxies. I am not responsible for any TOS violations or bans.The difference between global and custom prompts is simple and true to the name. A global prompt applies to all proxies you make. Custom prompts are only available through specific proxies you are using in that moment, so when you make a proxy, you will see "custom prompts" this is where you would input your commands, lore library, plugins, and scenarios. For example, for global prompt I have <FORCEMARKDOWN> but for custom prompt, for one proxy I have <REALISTICDIALOGUE> and for a different proxy I have <BETTERSPICE> <NPCNEEDS> <REALISTICDIALOGUE> <NOCLICHES>.