kayetanaYour AI platform.
In your own building.
kayetana brings chat, answers from your documents, search, transcription and custom models to a server in your office. No cloud, no data sharing, no per-token fees.
- Prompts are never stored
- Open models
- Works offline
Runs on open models you already know
- Qwen3
- gpt-oss
- bge-m3
- Whisper
- Ollama
Use AI without giving your data away.
Your data stays put
Models, documents and logs live on your server. Not a single prompt goes to a cloud provider.
Fixed costs
No per-token billing. Use kayetana as much as you like, across the whole team.
Quick to set up
One script installs models, services, HTTPS and backups. After that, it runs on its own.
Everything your team does with AI.
From chat to your own API: kayetana covers the everyday tasks and shows you right away in the Playground how it works.
Answers from your own files
Policies, contracts, handbooks: paste them in or upload PDF and Word. Every statement shows which source it came from.
Sort text automatically
Route tickets and emails to categories, with a few examples or a classifier you trained yourself.
Speech becomes text
Transcribe meetings and voice messages, with timestamps. MP3, M4A, WAV and more, up to 100 MB.
Your models, your choice
Chat, reasoning, search and speech with open models. Add new ones, set defaults, train your own.
Full visibility
Tokens and requests per day, key and model. Per-key limits prevent surprises.
How far does your data travel?
With cloud AI, every prompt goes to a data center and back. Enter your city and compare.
Up and running in three steps.
Set up the server
A Mac Studio in your office, quiet and smaller than a shoebox. All it needs is power and a network cable.
Set up models and team
We load the right models and set up HTTPS and backups. You invite your team.
Get going
In the browser through the Playground, or from your own apps through the API with your own keys.
Privacy is built in, not bolted on.
kayetana was built from day one to run on your own network. These principles apply with no extra setup.
No content stored
Prompts and answers are only passed through. kayetana records the time, model and token count.
Your retention rules
Usage logs and uploads are deleted automatically on your schedule.
Roles and permissions
Owners, admins and members only see what they should.
HTTPS on your network
Encrypted connections with your own certificate, even without internet access.
Nightly backups
The database and files are backed up every night and kept for 14 days.
Runs offline
You only need internet for model updates. Day-to-day operation is fully local.
One API for every task.
Chat, embed, rerank, classify, tokenize, transcription, datasets and fine-tuning through one REST API with bearer keys.
- Streaming with server-sent events
- Answers with sources through the documents parameter
- Monthly limits and usage per API key
- Working sample code in the Playground
curl https://kayetana.local/v2/chat \ -H "Authorization: Bearer $KAYETANA_API_KEY" \ -H "content-type: application/json" \ -d '{ "model": "qwen3:14b", "messages": [{ "role": "user", "content": "How much does the company cover for hotels in Hamburg?" }], "documents": [{ "id": "travel", "data": { "text": "Hotel costs are covered up to €140 per night, up to €170 in Munich, Hamburg and Frankfurt." } }] }'
import os, requests res = requests.post( "https://kayetana.local/v2/chat", headers={"Authorization": f"Bearer {os.environ['KAYETANA_API_KEY']}"}, json={ "model": "qwen3:14b", "messages": [{"role": "user", "content": "How much does the company cover for hotels in Hamburg?"}], "documents": [{"id": "travel", "data": {"text": "Hotel costs are covered up to €140 per night, up to €170 in Munich, Hamburg and Frankfurt."}}], }, ) print(res.json()["message"])
{
"message": {
"role": "assistant",
"content": [{ "type": "text", "text": "In Hamburg, hotel costs are covered up to €170 per night." }],
"citations": [{
"start": 0, "end": 56,
"sources": [{ "type": "document", "id": "travel" }]
}]
},
"usage": { "tokens": { "input_tokens": 186, "output_tokens": 21 } }
}
Get started in two minutes.
Short videos and step-by-step guides for every feature, in English and German.
For approved accounts only
One plan. Everything included.
No per-seat or per-token billing. You pay for the platform, not for every question.
kayetana Complete
Contact us for teams of any size
- Every feature included
- Unlimited users and requests
- On-site or remote setup
- Platform and model updates
- Nightly backups
- Email and phone support
- Tutorials and onboarding
- API at no extra cost
Hardware
Mac Studio your own, or supplied by us
- We recommend 36 GB of memory or more
- Quiet, efficient, fits on any desk
- More memory runs larger models
What hardware do we need?
A Mac Studio with Apple silicon. With 36 GB of memory it runs models up to about 20 billion parameters alongside search and transcription. More memory allows larger models.
Does kayetana work without internet?
Yes. You only need internet to download models and updates. Day to day, everything runs on your own network.
Which models are included?
Qwen3 in two sizes, gpt-oss 20b, bge-m3 for search and classification, and Whisper for speech. You can add more open models at any time.
Who can see our data?
Only you. Prompts and answers aren't stored; kayetana only records the time, model and token count. Uploads and logs are deleted on your schedule.
How do we connect our own apps?
Through the REST API with your own keys, each with a monthly limit. The Playground shows working code in Python, TypeScript and curl for every request.
What if the server fails?
kayetana backs up its database and files every night and keeps 14 days of backups. For real emergencies, we also recommend an external drive.
See kayetana in action.
In 30 minutes, we'll walk you through the platform using your own examples.