NOTE: This guide is no longer current. LiteRouter has removed all of the unmetered Claude LLMs.

🥚Claude 4.5 Guide!🥚
Super easy, takes about 5 minutes to set up!

Not free but actually affordable now! Rather than ~$8 USD for every 100 messages, its $20/mo for 100 messages per day at my settings. More or less if you adjust.

Note: I'm not affiliated with LiteRouter in any way, I've just payed out the ass for Claude on OpenRouter and I'm super thankful that this is now available, so I wanted to share with the class.

UPDATE: New changes to the Chat Optimization! Changes are highlighted in red.

Step 1:
Go to LiteRouter.com and create an account. Login with that account.

Note! You can actually stop right here and use the free models for up to 50 requests per day! But since we're talking about Claude let's continue on...

Step 2:
Go to Dashboard → Billing & Payments → Ultimate. That's the tier that has access to the Claude models. It's actually $17/mo US right now but it'll go up to $20/mo eventually. If you want to use Claude (again, there are free options that cost no money) then you'll have to subscribe at the Ultimate tier. (I used PayPal.)

You now have access to Claude 4.5 via LiteRouter. Sonnet is generally regarded as the one to use.

Step 3:
Generate your API key. Select API Keys → Create New API Key → Then Copy that API key.

Step 4:
Almost done on LiteRouter! Now we just need to make sure that Claude is able to actually remember stuff from your chat.
On the API Dashboard go to Chat Optimization. I'm currently using AI Response settings at None which is default.
(I think that this kinda works like an advanced prompt? But I'm not positive. You're free to experiment with this but since I'm using an AP I'm just leaving it off.)
Scroll down to the bottom and update the Chat Memory settings. You can see my settings that I use. The larger this is the better the chat memory!

Some notes about the way this works:
An Ultimate account gives you 5000 credits per day. Each request costs 1 credit. Different LLMs apply different modifiers.
(For the free LLMs that modifier is 0x, meaning each message costs zero credits no matter what.)
For the big boy LLMs, that modifier makes each request cost more credits. For instance, Claude Sonnet 4.5's modifier is 10.8x, so one request using that LLM on default settings costs 10.8 credits. That means you'd get about 462 requests per day with those settings. Increasing the Chat Memory (which you should do) adds an additional modifier.
With the settings I have specified, the math looks like this:
10.8 × 4.5 = 48/6
5000 ÷ 48.6 = 102.88
Therefore you'd get 102 requests per day with Claude Sonnet 4.5. It's pretty fast, and the messages are short compared to Gemini and GLM (closer to JLLM and Deepseek V3) so that might not be enough but it all depends on how much you chat and reroll. You can definitely squeeze more messages per day by dropping the conversation context and system context, just bear in mind that'll make the LLM more forgetful.
(That'll look like the LLM forgetting what characters are wearing, where they're standing, what they were doing, etc.)

UPDATE: LiteRouter has made changes to the Claude Chat Memory Settings since this guide was originally posted. The System Context and Conversation Context have been expanded! Thanks to Herki for bringing this to my attention! The settings I'm currently testing are pictured above, 7,500 System Context and 15,000 Conversation Context.

So what do these settings mean?

Per LiteRouter:

System Context is used to save system prompts, character definitions, lore books, and presets.
Conversation Context is used to save your conversation history with the AI.

Just what exactly this looks like as far as text handed back and forth between Claude I'm not sure. I think what it's trying to say is that if you're using high token bots with lorebooks and a long Custom Prompt, you may want to increase System Context. I think 7,500 tokens should be enough for the vast majority of bots.

If you're in a long roleplay with a bot, you may want to increase the Conversation Context. For my use, I think 15,000 tokens for chat history is somewhat pushing the bottom end of acceptable. You may want to increase this over time if you find the LLM being forgetful.

Step 5:
Now you're going to setup the proxy on JAI. This is just like every other proxy setup process, but just in case you're not familiar, here's how you do it.

Enter a chat and click the purple proxy button. Click "+ Add Configuration". Copy and paste the following values into their respective fields:

Configuration Name:
Claude Sonnet 4.5 LiteRouter

Model Name:
claude-sonnet-4.5
OR
claude-opus-4.5
(You're welcome to try Opus but I don't find a market increase in quality over Sonnet.)

Proxy URL:
https://api.literouter.com/v1/chat/completions

API Key:
[This is the API Key you generated in Step 3]

Custom Prompt:
Okay this is where it gets sorta complicated. There are so many of these you can use. You can even rawdawg it without an advanced prompt. I have two that I like to use. Brbiekiss's Universal Prompt (it's like a quarter of the way down in that document) and EgoAlter's Gemini Prompt. (Shoutout to EgoAlter!) Just copy one of those and paste it into Custom Prompt if you want. Feel free to experiment with these!

Make sure you SAVE CHANGES → SAVE SETTINGS before leaving this screen.

Finally change your generation settings to match mine. Temperature at 0.9, Max Tokens to 0 (no limit), and Context Size to match your Claude Total Context Window under Chat Optimization.

Step 6:
???

Step 7:
Profit!

And you're done! Enjoy roleplaying with arguably the best model out there!

Remember. There is no war in Ba Sing Se. We've always been at war with Eurasia. There are no Easter Eggs in this bot.

If this guide was helpful to you, maybe consider dropping me a follow? 👉👈🥺

And hey, while I have you here, why not chat one of my bots? How about your thicc best friend with a huge crush? Or your therapist who's decided to break the rules and go a little yandere on you? Or maybe you're a Tyler and you REALLY like cops? There's a lot more on my page and most of them are wholesome fluff, smuff, and comfort.