Your knowledge.
Your AI.
Your infrastructure.
KnowAI is a self-hosted knowledge-management system with built-in AI. Ingest documents, search them semantically, query them through chat – no cloud, no API bills, no data leaving your network.
Four ways to put your knowledge to work.
See KnowAI in action – each capability in a short clip, captured straight from the running application.
An assistant that thinks with you.
Brainstorm, draft and summarize in a free-form chat with the local language model – no token limits, and nothing ever leaves your premises.
Answers grounded in your own documents.
Upload manuals, reports and policies and query them conversationally – every answer with source citations and the matching figures from the original.
Talk to your database – no SQL required.
Connect PostgreSQL, MySQL, SQL Server or SQLite once and ask in plain language. KnowAI generates safe, read-only queries and visualizes the result automatically as a table or chart.
Turn spreadsheets into insight – in seconds.
Upload Excel or CSV files and ask for analyses in natural language. KnowAI picks the right chart type and returns an interactive Plotly chart ready to export.
Everything an internal knowledge system needs.
Twelve building blocks. On your hardware. Zero external dependencies.
Document chat
Upload documents (PDF, DOCX, TXT), index them, and query them through a RAG pipeline – with source citations.
Chat
Free-form chat with the local language model – no token limits – ideal for questions, drafts and ideas
Reports
Upload Excel and CSV files and generate interactive Plotly charts via natural language.
Database chat
Connect SQL databases (PostgreSQL, MySQL, SQL Server, SQLite) and query them in natural language – safe, read-only queries with automatic chart rendering.
Multi-user
Role model with Superadmin, Admin, Editor and User. SSO-ready on request.
Tagging
Categorize documents and reports with colored tags, then narrow down chat scope to a subset.
Self-hosted
Fully containerized via Docker Compose. Postgres, pgvector, Ollama, Supabase Studio – all on-prem.
OCR & image extraction
Scanned PDFs and images are recognized via OCR (German + English); embedded graphics appear directly in chat answers.
Hybrid retrieval & re-ranking
Semantic search, keyword matching, HyDE and cross-encoder re-ranking deliver more precise answers – with multi-turn context.
In-app model management
Download, update or delete Ollama models directly from the app. Live progress, no CLI.
Live system status
A status chip in the header shows component availability at any time; missing subsystems are automatically disabled in the UI.
Multilingual
The interface is available in German, English, French, Spanish and Italian. Each user sets their language individually – independent of the server installation.
Cloud AI or KnowAI?
The difference becomes obvious when you put both columns side by side.
Typical cloud AI provider
- Documents stored on third-party servers
- Monthly or token-based billing, hard to predict
- GDPR data-processing agreement per vendor
- Model selection limited to vendor's catalog
- Service outage = your operations stop
KnowAI on your own infrastructure
- Data physically stays in your data center
- One-time price – €699 per install, that's it
- No DPA needed because no transfer happens
- Pick any model: Llama, Mistral, Gemma, Qwen, …
- Keeps running offline, in air-gapped networks too
One price. One install. No surprises.
No subscription, no per-seat fee, no hidden costs. Buy once, use as long as you like.
What a self-built RAG system really costs
A typical starter RAG eats up five-figure setup costs plus recurring monthly fees. KnowAI costs €699 once – with no monthly fees and more features.
Typical starter RAG
- €8,000–15,000 one-time (setup)
- €200–400 per month, recurring
- ≈ €2,400–4,800 extra per year
KnowAI
- €699 one-time – that's it
- €0 per month – no subscriptions
- More than a starter RAG: OCR, reports, multi-user, hybrid retrieval & more
And most importantly: KnowAI is not a plain starter RAG. You additionally get OCR & image extraction, interactive reports, a multi-user system, tag filters, in-app model management and hybrid retrieval with re-ranking – see all features. More capability at a fraction of the cost.
- Unlimited internal users
- All features, no feature-gating
- License key delivered via e-mail
- Install guide
- Priority bug fixes
- On-site install & onboarding
- Training for your internal admins
- Custom integrations & adjustments
Frequently asked questions
What happens to my documents?
They are indexed in a PostgreSQL database with pgvector – on the same machine you install KnowAI on. No data is transferred to third parties.
Do I need a GPU?
No, KnowAI runs CPU-only too. With an NVIDIA GPU answers come back significantly faster. A GPU setup guide is included.
Can I freely choose the language model?
Yes. Llama 3.2 is ready to use out of the box. Further open-source models (e.g. Mistral, Gemma 2, Qwen) can be downloaded directly within KnowAI.
What's included in the €699 one-time price?
The software license for one installation in your infrastructure. Support is optional.
How is the license key delivered?
Right after a successful Stripe payment you automatically receive an e-mail with the cryptographically signed license key at the address you entered in checkout.
Can I try the software before buying?
Yes. The installer ships with a 30-day trial license at full functionality. No account, no credit card, no auto-renew.
Can I alternatively use cloud-hosted language models?
Yes. KnowAI can be configured to use an external endpoint (your own cloud server or a compatible API provider) instead of the local Ollama instance. Important: as soon as data leaves your infrastructure, this step needs to be aligned with your company's data-protection policy – particularly regarding GDPR, data-processing agreements and any cross-border data transfers.
Where is the software installed within my company?
The software is installed on your own local server or cloud infrastructure. KnowAI is fully self-hosted and integrates easily into your existing systems.
What are KnowAI's system requirements?
- Minimum: 4 CPU cores, 16 GB RAM, 50 GB free disk space
- Recommended: 8+ CPU cores, 32 GB RAM, 100+ GB free disk space, NVIDIA GPU with 8+ GB VRAM (optional)
These values refer to the standard installation. Depending on the Ollama model you pull yourself (e.g. Mistral, Qwen or Llama 3.1), RAM and VRAM requirements may be higher. When using cloud-hosted models, the compute-heavy local inference is offloaded – a local GPU no longer provides any benefit and RAM requirements drop; however, the database, vector store and embeddings still run locally.
How do I update to a new version?
Download the latest installer from the download page and run it. The system detects your existing installation and updates all components automatically – your data and settings are preserved.
Your knowledge. Your AI.
KnowAI is a self-hosted knowledge-management system with built-in AI. Ingest documents, search them semantically, query them through chat – no cloud, no API bills, no data leaving your network.