Two models
Kayts Mihr answers fast for everyday tasks. Kayts Anahit, our smartest model, takes on complex work.
Early access
An assistant that drafts your letters
Sign in to start your first conversation.
or
Join the waitlist
Leave your email and we'll write once, when your spot is ready.
You're on the list
Check your inbox
- Data stored in the EU
- Opt out of training anytime
- Cancel anytime
Features
Everything you need in one chat
Write, ask, show a photo or add a file. Kayts works with whatever you bring.
Thinking mode
For harder questions, let it think the problem through before it answers.
Images
Add up to four photos or screenshots to a message and ask about what's in them.
Files
Attach notes, tables or code as text files. Kayts reads them along with your question.
Dictation
Speak instead of typing, in the language you choose.
Answers your way
Tell Kayts about yourself and pick the length and tone of its answers.
Research
Chat with your papers
Search over 250 million scholarly works, paste a ResearchGate link or a DOI, or add your own files. Kayts answers from your sources and shows where each claim comes from.
- 250M+ works
- ResearchGate
- DOI
- Your files
- Searches the OpenAlex catalog with filters for year, type, topic and open access, sorted by relevance, citations or date
- Adds a paper in one click and fetches its full text when it's open access; otherwise it works from the abstract
- Reads PDF, Word, Excel, CSV and text files right in your browser
- Cites its sources like [1] and says when they don't cover your question
- In any chat, looks up papers on its own when you ask about research, and lists the ones it found
Up to 12 sources in one chat. Scanned PDFs are read with text recognition (OCR), up to 30 pages; the result may contain errors.
deep residual learning
Deep Residual Learning for Image Recognition
OA
1
Why do residual connections help very deep networks?
Plain networks get harder to train as they grow deeper: past a point, even training accuracy drops.
Residual blocks learn only a change to their input, which is added back through a shortcut. That made networks with over 100 layers train well. 1
y = ℱ(x, {Wi}) + x
Privacy
Your conversations stay yours
Kayts runs on open-source language models that we run ourselves.
Stored in the EU
Your account and chats are kept in the European Union.
You decide about training
Your chats can help train future Kayts models. One switch turns it off, when you sign up or anytime in settings. For users under 18 it stays off unless they turn it on.
Photos without location
Images are resized in your browser before upload, which also removes hidden data such as GPS location.
Files read on your device
Research files are opened in your browser. Only their text goes with your questions, never the file itself.
Dictation uses your browser's speech recognition: in Chrome and Edge, your voice is processed by Google or Microsoft.
Plans
Start free. Go Pro when you need more.
Both plans get the same assistant. Pro gives you Kayts Anahit, our smartest model, more room and extra tools.
Free
Try it out. No card needed.
$0 / month
Free forever
Start for free- Daily and weekly usage limits
- Kayts Mihr, our fast model
- Try Kayts Anahit every day
- Images and file uploads
- Unlimited chat history
- Basic / commands
Pro
For people who work with it every day.
$12.99 / month
Opens with the beta
- Everything in Free, plus:
- Kayts Anahit, our smartest model
- Much higher usage limits
- Full memory
- Larger files
- Priority at busy times
- All / commands
Pro opens with the beta. Until then, everyone uses Free.
FAQ
Frequently asked questions
What is Kayts?
An AI assistant that writes, explains and plans with you. It's built on open-source language models that we run ourselves.
Which models can I use?
Two: Kayts Mihr, our fast model for everyday tasks, and Kayts Anahit, our smartest model for complex work. They are named after Armenian deities: Mihr, god of the sun and light, and Anahit, goddess of wisdom. Both read images and can think a question through before they answer. Free includes Kayts Mihr and a daily taste of Kayts Anahit; Pro includes Kayts Anahit.
Is there a free plan?
Yes. The Free plan costs nothing and needs no card. Sign in with Google or your email and start chatting. Apple, Microsoft and ORCID sign-in are coming soon.
How are the plans different?
Free gives you Kayts Mihr and a daily taste of Kayts Anahit. Pro, for $12.99 a month, gives you full access to Kayts Anahit, much higher limits, memory, larger files and priority at busy times.
What happens when I reach my limit?
Like Claude, Kayts has two usage windows: a 24-hour one and a 7-day one, each starting with your first message. The daily allowance lets a busy day be busy; the weekly one is about three such days. When your Kayts Anahit allowance runs out, the chat switches to Kayts Mihr, so you can keep going. If you regularly need more, Pro will give you much more room once it opens.
Are my conversations used to train the model?
Not yet: fine-tuning on conversations starts in a later stage of development, and then only while “Help improve Kayts” is on. It's on by default for adults: you can turn it off when you sign up or anytime in settings, and from then on your chats aren't used. For users under 18 it stays off unless they turn it on.
Where do the papers come from?
From OpenAlex, an open catalog of over 250 million scholarly works. Kayts finds a paper by search, DOI or ResearchGate link and fetches its full text when it's open access. Some publishers block downloads: then Kayts works from the abstract, and you can add the PDF yourself. Pages on ResearchGate itself are never downloaded.
What does Kayts remember about me?
On the Pro plan, it can remember facts you share, like your job or how you like answers written. You can see and delete everything it remembers.
Can I cancel anytime?
Yes. Your plan stays active until the end of the period you've already paid for.
Which languages does it speak?
Kayts understands over 200 languages and dialects and answers in the one you write in: English, Spanish, German, French, Polish, Ukrainian, Arabic, Hindi, Japanese and many more. It's strongest in English and other widely spoken languages; in rarer ones, answers may be less fluent.