← Back to home
Before you ask

Questions, answered.

Quillin’ is invite-only right now: a solo side project, shared carefully. Here's what most people want to know before they request access.

What does "bring your own key" (BYOK) mean?

You connect Quillin’ to an AI you already have (Claude, GPT, Gemini, Grok, Mistral, Groq, or local Ollama) using your own API key. Quillin’ sends your text directly to that provider and gets the rewrite back. There's no Quillin’ AI, no Quillin’ bill, and no Quillin’ server in the middle.

Why invite-only?

Quillin’ assumes you know what an API key is and are comfortable with BYOK. Most people aren't, and that's fine. Keeping it invite-only means the people using it are the right fit for it right now.

How long does an invite take?

I send invites in batches and check the request inbox every few days, so you won't hear back instantly, but you're not shouting into a void either. Request an invite once and sit tight. It's a solo project, so it moves at solo-project speed.

How do I get an API key?

Gemini and Groq both have free tiers generous enough for daily use. Go to aistudio.google.com for a Gemini key or console.groq.com for Groq. Takes about two minutes.

How much does it actually cost to run?

Gemini and Groq have free tiers generous enough that plenty of people never pay a cent. On paid models you're billed per token by your provider, not a flat Quillin’ fee, so the bill tracks your own usage and model choice. And going feral is the whole point: firing a stressful HR letter through Unhinged, Pirate, and Victorian Offended until you're laughing instead of spiraling is exactly what Quillin’ is for. Each rewrite is a few more tokens, so if that's your daily ritual, keep an eye on the meter. Your provider shows exact usage in their dashboard, and most let you set a hard spending cap so it can never surprise you. For what it's worth, that kind of goblin use has rarely topped about $7 a month for me. But you're the one holding the dial.

Are my API keys safe?

Yes. Keys live in Chrome's local browser storage and never go to any Quillin’ server, because there is no Quillin’ server. When you run a check, your browser makes a direct request to your AI provider using your key. Quillin’ is just the interface.

Does Quillin’ store or log my text?

No. Your text goes from your browser directly to the AI provider you've chosen, under their privacy policy. Quillin’ doesn't see it, log it, or touch it in transit.

Does my AI provider train on my text?

That's up to the provider, not Quillin’. Your text goes straight to them under their terms, so their policy decides whether it's used for training. Most paid API tiers don't train on your inputs, and if you run Ollama locally, nothing leaves your machine at all. If it matters to you, check your provider's API data policy, or go fully local.

The quill didn't appear. Now what?

Highlight your text, right-click, and choose "I be Quillin’ this." You get the same rewrites. If that's also missing, reload the tab. Chrome only injects the extension into tabs opened after install.

Which browsers work?

Chrome and Brave today. A native Mac app is in the works (no browser needed). Windows if enough people ask nicely. Let me know if you want to be notified.

Does it work on my phone?

Not yet. Quillin’ is a desktop browser extension for Chrome and Brave. Mobile browsers don't support extensions the same way, so there's nothing to install on a phone right now. A native Mac app is in the works, and mobile could follow if enough people want it.

Can I use Ollama for totally local, private AI?

Yes. Download Ollama from ollama.com, pull a model (ollama pull llama3.1), and enable it in Quillin’ Settings. No API key needed, no data leaves your machine.

What are standing instructions?

Rules that apply to every check, silently and automatically. Examples: Never use em dashes, Oxford comma always, Canadian spelling. Set once in Settings, forget about it forever.

How many tones can I use at once?

Five at a time. You choose your five defaults in Settings, and every right-click runs all five side by side so you can compare. Five is just how many appear at once, not a limit on what you can build. The styles themselves are limited only by your imagination.

Can I keep tweaking a result?

Yes, and this is where it gets fun. Once the rewrites come back, keep nudging whichever one you like. Start with Pirate, then give him an Italian accent, make him a mama's boy, add a few years living in Germany and a Japanese wife. It keeps refining until it sounds exactly right.

How many languages can it translate?

Quillin’ has a long list built in, but the real answer is however many your AI can handle. Translation happens on your provider's model, so its range is the ceiling, not ours. Google's Gemini covers the most by a wide margin, so if translation is your main thing, start there. Emoji included.

Is there a limit on how much text I can send?

Quillin’ doesn't cap it. The ceiling is your model's context window and whatever your provider allows per request. Short selections and normal paragraphs are no problem. Paste in something enormous and larger models will handle more, though you'll pay for more tokens.

Will it ever cost money?

Free for now. I reserve the right to change that as it grows. Want to be notified if a paid option launches? Drop me a note.

Is there support?

It's a solo side project, shared as is. There's no support desk, but bug reports and questions get a real response. Use the contact page.

The model explained

Bring your own key: what that actually means.

Most AI writing tools charge a monthly subscription and run your text through their own servers. Quillin’ doesn't. Instead, you connect the AI you already use (Claude, GPT, Gemini, whatever). You get the key directly from the provider. Quillin’ is the interface. The AI is yours. No middleman, no markup, no Quillin’ bill.

Why it works in your favour
You pay the AI provider directly. API pricing is per token, not a flat monthly fee, so you only pay for what you actually use. Light use costs little. Heavy use on a premium model costs more. Either way, there's no Quillin’ markup on top.
Your text never touches a Quillin’ server. When you run a check, your browser sends the request straight to your AI provider. Quillin’ has no server to intercept it, log it, or leak it.
You choose the model and can switch anytime. Use Claude for nuanced rewrites, Gemini when you want speed, Groq when you want free. Change it in settings in ten seconds.
Free tiers exist and are genuinely useful. Gemini and Groq both have free tiers with limits generous enough for real daily use. You can run Quillin’ at zero cost.
Fully local is an option. If you run Ollama on your machine, no text leaves your computer at all. No API key, no internet, no data anywhere.
!
The honest trade-offs
Setup takes a few minutes. You need to create an account with an AI provider and paste an API key into Quillin’ settings. It's a one-time thing, but it is a thing.
You're responsible for your own key. API keys are like passwords. If someone gets yours, they can run up charges on your account. Keep it out of shared docs and public repos.
Costs can vary. Paid models charge per token, and more capable models cost more per request. It's rarely significant, but it's not literally zero for everyone.
If the AI provider is down, Quillin’ can't check. Quillin’ doesn't have a fallback AI of its own. Outages are rare, but they happen upstream and there's nothing to do but wait.
It's not for everyone. BYOK assumes you're comfortable with API keys and a little setup. If that sounds like a headache, it probably is. And that's okay.

Not sure which provider to start with? Gemini has a free tier at aistudio.google.com and takes about two minutes to set up. If you want something even faster, Groq is free at console.groq.com. Either one works on day one with no credit card.

See how to add a key →

Ready to try it?

Invite-only right now. If BYOK is your kind of thing, request access and I'll add you to the list.