The self-hosted AI platform, on your own servers.
Use AI on your company knowledge without your data leaving the building. RAGSuite runs on your own servers, with AI search, an AI chatbot and a source on every answer. With no internet connection if you need.
No servers of your own? RAGSuite Engine in Germany →
:9191 · :9090PostgreSQL · ChromaDBOllama · vLLMRuns on your servers in your own network. The model runs locally or via API, you decide per project.
What does on-premise AI mean?
On-premise means the software runs in your own environment, in your data centre, on rented servers you control. For on-premise AI that applies to everything that touches your data: documents, search index, history and, if you want, the language model itself. Nobody but you has access.
Why companies self-host their AI.
Data protection without compromise.
Personal data, contracts and internal policies stay in your environment. That makes talks with data protection and the works council easier.
Predictable costs.
No price per request. The Community Edition is free; Enterprise comes at a fixed price by agreement.
No dependency.
The core is open source. You decide which model to use and can switch at any time.
Works without internet.
With a local language model, RAGSuite runs fully offline (air-gapped), for example in sensitive networks.
Which deployment model fits you?
Three ways to self-host AI: pick the one that fits your network.
Rather not run servers?
Then we run RAGSuite for you, in RAGSuite Engine in a data centre in Germany, with the same features.
What runs on your servers.
Every part of RAGSuite runs on your side. None of it calls us.
FastAPIReactPostgreSQLRedisChromaDBOllama · vLLMRun a local LLM, or use a model via API?
Both work, and you decide per project. A local language model keeps everything in-house and works offline; you need servers with a suitable GPU. A model via API needs no GPU of your own; only the question and the matching passages go to the provider you choose. Documents, index and history stay with you either way.
Nothing leaves the building. Works with no internet connection (air-gapped).
Data leaving: noneOnly the question and the matching passages go to the provider you chose, with your own key.
Data leaving: question + passages → EU providerOnly the question and the matching passages go to that provider. Your documents stay with you.
Data leaving: question + passages → model providerRAGSuite itself never sends data to us. No phone-home, no telemetry. Compare models →
Three ways compared: AI services from abroad, RAGSuite Engine, on-premise.
| Criterion | Foreign cloud AI service | RAGSuite Engine | RAGSuite On-Premise |
|---|---|---|---|
| Where do documents and index live? | With the provider, often outside the EU | Data centre in Germany | On your servers |
| Who runs the servers? | The provider | RAGSuite | Your IT team or a partner |
| Without internet? | No | No | ✓ Yes, with a local model |
| Costs | Often usage-based | Fixed plan price from €39 a month, credits included | Community free, Enterprise by agreement |
| Code inspectable? | Usually no | Core open source | ✓ Yes, Apache 2.0 |
| Choice of model | Set by the provider | Choose per project | Free, local too |
On-premise also means you run the servers. Without your own IT team, use RAGSuite Engine or let a partner run it.
What data protection and the works council want to know.
- No telemetry: RAGSuite sends no usage data, neither to us nor to third parties.
- Deletion periods with receipts: old data is deleted automatically, and every deletion gets a receipt.
- Storage can be switched off: conversation storage can be turned off for chat and search.
- AI notice in the widget: users see they’re talking to an AI.
- Audit log: the last 15 days in the Community Edition, full history with export in Enterprise.
- DPA: for Enterprise customers whenever support involves data access.
AI data protection →Trust Center →DPIA template →AI and the works council →
Is your company ready for AI on its own data?
- Which data may go into the index?
- What data protection and the works council need up front
- Local model or API: the deciding questions
We only use your details to reply to you. Privacy policy
The licence doesn’t need the internet either.
The Community Edition needs no key. You activate Enterprise with a signed offline key that is checked on your server: no licence server, no remote kill switch.
$ ragsuite activate \ --key offline.key \ --bundle ragsuite-ee-<version>.encbundle \ --restart ✔ licence checked locally · no licence server
What self-hosted AI costs: honestly.
The Community Edition costs nothing. Enterprise for running on your own servers comes at a fixed price by agreement: depending on users, operations and support. Add your servers and, if you run a local language model, a suitable GPU. With a model via API you pay the provider directly.
Onboarding, SLA and support for air-gapped environments are available by agreement.
Questions about on-premise AI.
What is on-premise AI for companies?
On-premise AI means an AI application runs entirely in your own environment: in your data centre, on servers you control or with a host of your choice. Documents, search index and history never leave the building. With a local language model, even answer generation happens on your side.
Do we need a GPU for on-premise AI?
Only if you run the language model yourself. The RAGSuite platform runs on standard server hardware. If you use a model via API, you need no GPU of your own. For a local model, the GPU you need depends on the model size; we'll advise you.
Can RAGSuite run completely without internet?
Yes. With a local language model, for example via Ollama, RAGSuite runs air-gapped, with no outbound connection at all. The Enterprise licence is activated offline too, with a signed key checked on your server. This suits networks where no outbound connection is allowed.
How much effort is it to run?
Installation takes three CLI commands, native or Docker. Updates also run via the CLI and keep your database and configuration. Your IT team runs the servers; if you don't have one, partners can run it for you, or you use RAGSuite Engine in Germany.
On-premise or RAGSuite Engine, which fits better?
On-premise fits if you have your own IT, process very sensitive data or must work without internet. RAGSuite Engine fits if you want to start fast and run no servers; it runs in a data centre in Germany. Both use the same platform, and you can switch later.
Is on-premise AI automatically GDPR-compliant?
Not automatically: GDPR also covers your processes. But on-premise removes a big hurdle: data doesn't go to third parties. RAGSuite supports you with deletion periods and receipts, switchable storage, an audit log and a DPA for Enterprise customers. Responsibility for the processing stays with you.
Self-hosted AI, on a setup like yours.
We’ll show you RAGSuite in an environment like yours and work out which deployment model fits.