Frequently asked questions
On basebox, prices, security and use in hospitals and administration.
General
An AI platform for hospitals and the public sector. Your staff work with AI on your own data, in our cloud in a German data centre or on your server.
Many platforms are an access point to models of other providers, such as ChatGPT or Gemini, even if they are hosted in Germany. By default, basebox does not send your data to any AI model of another provider: the model runs on our servers in Germany or on a server dedicated to you. An external model is only available as a separate app, if you explicitly have it set up. That is why patient data may go in without you anonymising it first. On top of that come permissions from your user directory, sources for every answer and connectors to hospital systems such as OrbisF1 and roXtra. Why anonymisation is not enough →
Copilot can stay. basebox is added for sensitive data such as patient, citizen and social data that your organisation wants to process in Germany and under its own control. It will not become a second major project: sign-in runs via your existing user directory, including Microsoft Entra ID, the apps are ready, and in the cloud you are working after 2 minutes. The apps show your staff which tasks belong in basebox.
AI, features, apps, connectors and security are the same. In the cloud, basebox runs on our servers in Germany, we operate everything, you pay per person per year and start straight away. With the server, basebox runs on hardware dedicated to you, on your premises (without an internet connection if you wish) or in a data centre, by default at noris network in Munich. You buy the server once and pay a licence per server; anyone may sign in. From about 230 people, the server is cheaper over five years. Calculate for your number →
Yes. basebox grows without a new system: apps, permissions, connections and logs stay the same; only the computing power grows. With the server, on larger hardware that you procure yourself, for example from HPE or Dell; in the cloud, without any effort from your IT. Ready-made basebox servers are available up to 2,000 people, cloud list prices up to 20,000 people; the licence for larger hardware is in the calculator. For larger organisations, networks and groups, we clarify the hardware with you beforehand. Calculate with your number →
Yes. The data stays on your server or in your own environment in the German data centre and is never used for training. Data protection →
Yes, basebox is not tied to any model. By default, you work with the model we select and test for your tasks. On request, we enable further models for your organisation; your administrator decides who may use them. Each model shows where it is processed. Models in basebox →
Usually not, and with basebox you do not need it: your data does not leave your boundaries, so nothing has to be made unrecognisable. Why anonymisation is not enough →
No. We set up basebox, connect it to your IT and take over operation on request. Services and prices →
The location and the control. basebox runs on your server or in your own environment in Germany, which is why patient, citizen and social data may go in. Every action is assigned to a person and logged, and the permissions come from your user directory. Platform, model, computing power →
In the cloud, you start working immediately. The server from our hardware partner is delivered three weeks after the order. On existing hardware, we install basebox in one day, plus coordination with your IT, for example for VPN access. Connections to specialist systems follow afterwards. Rollout schedule →
Yes. You register free of charge and work with 1 million tokens in your own account in our cloud, with all features. On request, we show basebox in a meeting with your own sample documents. Request a meeting →
Features
Open-source models that we run ourselves: in the cloud on our servers in Germany, with the server on your hardware. We do not develop our own model. We install the best, most capable models that suit the available hardware, including on our own; there are no feature tiers. On request, we also set up an app that works with an external model, for example from Google or Anthropic, processed in Frankfurt. Your organisation can also connect interfaces of other providers. What this means for data protection →
We replace the AI model in basebox when a clearly better one appears, after our tests with tasks from hospitals and administration. That does not happen all the time. We update the external model in its own app within a few days once a new one is available and tested. For this we use Google Vertex in Frankfurt: no training with your inputs, logging by Google for abuse detection for at most 30 days. With other providers, such as STACKIT, inputs are not stored at all. If you want to connect other providers such as OpenAI, that is possible. We then point out the consequences for data protection and security.
Yes. Web search is a feature of basebox, not of the model. If your administrator enables it, it works with any model. Web search never sees your internal data. Security of web search →
Ja. Für sensible Daten arbeiten alle mit dem KI-Modell in basebox. Zusätzlich richten wir Apps mit einem externen Modell ein, etwa von Anthropic, OpenAI oder Google. Ihr Administrator legt fest, welche Personen und Gruppen eine solche App nutzen und welche Dokumente und Konnektoren sie bekommt. Das geht in der Cloud und beim Server, wenn Ihre Firewall die Verbindung zum Anbieter freigibt. Externe Modelle rechnen nach Verbrauch ab; bestellen Sie die Nutzung in der Lizenz mit, sind die Kosten im vereinbarten Umfang enthalten. KI hybrid nutzen →
Yes, on the server. Your developers keep their tools in VS Code or in the terminal, such as Cline, Kilo Code or Aider, and connect them via the OpenAI-compatible interface of basebox. A coding model then runs on your GPUs alongside the AI model for the whole organisation; the code does not leave the building. For this, the server needs room for that model; we work out with you how many developers can work at the same time. In the cloud, the quota is enough for questions about code, not for agents that program autonomously. What works and what does not →
You describe a task in your own words, and AI agents turn it into an app that handles it from then on, including across several steps and systems. The app works only with the permissions of the person who starts it, and write actions run only after approval by a human. From November 2026.
Yes, without programming. An app is a saved task with an instruction, knowledge source and model. Apps can be shared with groups or the whole organisation. Use cases →
PDF, Word, Excel, PowerPoint, text and CSV, plus images, scanned documents via text recognition and audio via speech recognition.
Yes, a REST API and an OpenAI-compatible API with a separate key per person. With these, you build basebox into your own applications. Interface for research and development → Docs ↗
These interfaces connect staff with an AI model. A whole organisation needs more: permissions from your user directory, apps that can be shared without sharing data, connectors to your systems with each person's permissions, a log, evidence such as a DPA and an agreement under Section 203 of the German Criminal Code (StGB), updates and support. basebox comes with all of this ready; your IT does not have to build and maintain it. You can keep using existing hardware for it.
No. vLLM is open-source software that runs an AI model on GPUs, and basebox uses it for that. basebox is the platform around it: sign-in via your user directory, permissions and groups, apps, connectors, log, updates and support. Docs ↗
No. basebox is a finished platform with licence, support and updates. It is built on open-source components, such as vLLM for the AI model and Keycloak for sign-in. The software bill of materials (SBOM) lists which components are included. Documents →
You keep working. If basebox becomes insolvent, the code will be published as open source so that you can continue to run basebox. On the server, basebox runs on your hardware anyway.
Prices
No, there is no limit per person, group or model yet. The cloud includes 20 million tokens per person per year, around 12 million words. The allowance is shared by your whole organisation, for all models; whoever uses little leaves more for others. Your administrator can see usage at any time in the dashboard. On the server there is no allowance.
Additional tokens cost €49 per million, with no difference between question and answer. You add them on request and they are activated within minutes. Nothing is ever bought automatically. In practice the allowance is more than enough: 50 people together have 1 billion tokens a year.
Yes, all prices are net plus VAT. Hardware prices are list prices of our hardware partner FAST LTA.
All features, support, updates, all apps and all connectors from our library. There are no feature tiers. All features → All connectors →
In the cloud, we add more people immediately. With the server, nothing changes; the licence covers the server. If its capacity is no longer enough, you switch to the next server and pay only the difference in the licence. The servers →
Per system between €0 (already in our library) and €13,800 for complex systems. The design is free of charge; you get a fixed price. Connectors and prices →
20 million tokens per person per year, around 12 million words. The quota applies to your whole organisation together. With 100 people, 2 billion tokens are available, no matter who uses them. Anything not used by the end of the contract year expires.
It evens out. There is no limit per person, only the shared quota of the organisation. If it is not enough overall, you book an add-on package at a fixed price. With the server, there is no quota.
Yes, at any time, flat rate €4,200. The servers →
By invoice. In the cloud, you pay per person per year; the minimum term is one year, and longer terms are cheaper. We agree payment terms in the quote. Calculate the price →
In the cloud €107,214 per year, i.e. around €214 per person. With a server: €109,643 once for the server, plus €21,569 licence and €9,868 service contract per year. All features, support and updates are included; list prices net. Calculate for your number →
In the cloud, one year; longer terms are cheaper. With the server, the service contract for the hardware runs for five years; you choose the licence for between one and five years or buy it once. After that, the licence renews for twelve months at a time unless you cancel three months before the end of the term.
Your data is with you and stays with you. In the cloud, you export your knowledge bases and settings, and then we delete your environment. With the server, the server belongs to you anyway.
In the cloud, per person using the AI. For a group or network of organisations, you conclude a joint contract; the people of all organisations count together and lower the price per person. With the server, you pay a licence per server. Anyone may sign in; we do not count accounts.
Yes. The licence agreement, DPA, TOMs, the list of sub-processors, the agreement under Section 203 of the German Criminal Code (StGB) and the annexes on support and operation are available for download (in German). You receive the price list with your quote. Licence agreement and annexes →
Security and data protection
In Germany, in the data centre in Munich East. The servers belong to our hardware partner FAST LTA; basebox uses them. We use additional computing power where it is available and where you want it: with providers in Germany and the EU such as STACKIT or Mistral AI, or with your own provider, for example through an existing framework agreement with Microsoft Azure. We only use offerings that do not train on your data. Which providers work for you is listed in the list of sub-processors.
Yes. It is an annex to the DPA. The basic infrastructure of the cloud is in Germany. If we use additional computing power for you, for example at STACKIT, Mistral AI or a provider of your choice, that provider is on the list. If you enable web search, Staan.ai with a data centre in the EU is added. If you use an app with an external model, Google Vertex with the data centre in Frankfurt is added. List of sub-processors (PDF)
An external model is only available as a separate app, in addition to the AI model in basebox. It does not run at basebox but at Google Vertex in Frankfurt. Google does not use your inputs for training and logs them for abuse detection for at most 30 days. Google is then a sub-processor of basebox, and because Google is a US company, the US CLOUD Act applies. That is why you decide with your data protection officer whether you use the app and for which data. Your administrator shares it only with the groups that are to use it. Everything else continues to run with the AI model in basebox, in Germany.
Not for your data. The AI model in basebox runs on our servers or yours without an outbound connection. What goes out is decided by the platform alone, and it releases nothing without your approval. If your organisation has requirements on origin, we implement them.
Yes. In the cloud, the model runs on our servers; we never train it with your data. With the server, everything runs on your hardware; nobody can train there.
That you use basebox as a processor, for what purpose and where the data is processed. We provide the details about basebox as a text module for your privacy notice.
No. basebox makes suggestions; people make decisions. For every answer, you see the sources and check them before you adopt anything.
We back up your environment daily and encrypted in a second data centre in Germany and keep the backups for 30 days. After the contract ends, you can export your data for 30 days; then we delete your environment. The last backups expire 30 days later.
Yes. Send us your template and we will review it. Or you use our DPA, which you can view in advance. View the DPA →
Because basebox is not a medical device. ISO 13485 applies to manufacturers of medical devices. For a platform, information security is what counts, covered by ISO 27001 and, for the cloud, C5 (Type 1 in progress). If you build a medical device yourself, we provide you with the information of a supplier. ISO 13485 and medical devices →
The hospital, as the operator, fulfils the B3S for hospitals. basebox provides the evidence for it: ISO 27001 certificate, TOMs and the description of architecture and data flows. A mapping of our measures to the B3S requirements is available on request. NIS2, KRITIS, B3S →
Implementation is under way. The software bill of materials (SBOM) is generated automatically, and the reporting channel for vulnerabilities is in place. The EU declaration of conformity will follow before the deadline on 11 December 2027. Cyber Resilience Act →
With the server, on your premises. In the cloud, in your own environment in our data centre in Germany. Additional services you enable, such as web search, run with providers in the EU.
No, never.
No. basebox is a German company, and the models run within your boundaries. It is different with models from US providers: the CLOUD Act applies to them even in European data centres. Data flow →
With the server, no: no telemetry, no usage statistics, no outbound licence checks. In the cloud, we only record operational data such as load and error messages, no content.
For AI with health or social data, usually yes; your data protection officer decides. As the controller, you prepare the impact assessment. We provide a template in which all details about basebox are already filled in. Download the template →
Yes, the DPA under Art. 28 GDPR is automatically part of the contract. DPA →
A dedicated package with answers on logging, permissions and co-determination, plus a model agreement as a template. For works and staff councils → Template and process →
basebox is a platform, not a high-risk system within the meaning of the AI Act. The obligations for your applications depend on how they are used. basebox provides the foundations for this: logging, approval by a human, documentation of the models and training on the AI Act. EU AI Act →
Yes. New models arrive by update; on request, we deploy a specific model, including your own. Your apps, permissions and data stay the same.
An open standard through which AI systems access data and tools. basebox uses it to connect your systems. Every access runs with the permissions of the person currently asking, read-only by default. All connectors →
No. basebox is a platform for general tasks such as drafting letters or searching guidelines, without a medical intended purpose. If your organisation builds an application with a medical intended purpose on it, your organisation is its manufacturer. For your technical documentation, we provide the information of a supplier. Is an AI platform a medical device? →
Yes. For Protestant organisations and Diakonie, the DPA under Art. 28 GDPR is sufficient (Section 30 DSG-EKD). With Catholic organisations, including Caritas, we additionally agree on the application of the KDG (Section 22 KDG-DVO). Our ISO 27001 certificate serves you as evidence under Section 9(7) KDG-DVO. Church data protection →
Cloud
Nothing. You register with your work e-mail address and work with 1 million tokens, with no time limit. You only book the licence when you want to continue.
Yes. Apps, knowledge bases and settings remain when you book the licence. You do not register again.
Technically yes, for one to five people and with weaker answers. We do not offer it, because the AI does not deliver there what an organisation needs in everyday work. For small teams, there is the cloud: from 5 people, without your own hardware, ready in 2 minutes.
Yes, as white label with your name and your logo, as operation for third parties. Each of your customers gets its own separate environment with its own user directory. You operate it yourself or we do, in our cloud in Germany or in your data centre. White label and operation for third parties →
Server
What matters is not how many accounts you set up but how many people ask questions at the same time. That is why a smaller server than the number of accounts suggests is often enough. In the configurator, we suggest the right server; before the quote, we calculate it with your figures. Calculate with the configurator →
Yes, if it has enough capacity for your number of people. We check it free of charge beforehand and assign it, based on its capacity, to one of our servers or a larger configuration; the licence depends on that. If you already have a server with GPUs, basebox runs on it. If the capacity is not enough, you start in the cloud and switch later. Calculate the price →
Yes. Your own platform remains a software project: every new feature, every new model and every piece of evidence is commissioned separately. With basebox, updates, models and features come with the licence. You keep using existing servers with GPUs; basebox runs on them. We discuss with you what the switch looks like for your organisation and what it costs. Contact →
When many people work at the same time, for example when the server is used at the upper end of its guideline range. Ongoing conversations and long documents then stay in memory and do not have to be recalculated. How many answers are produced at the same time is determined by the GPUs.
Support on working days from 9 am to 5 pm, first response within one working day, prioritised by severity. We do not commit to a fixed availability. Further service levels are available on request. With the server on your premises, availability also depends on your infrastructure. For the hardware, the service contract with FAST LTA applies.
Repairs and the replacement of defective parts, including expensive ones such as the GPUs. How quickly parts are replaced is governed by the agreed response time. That is why you usually do not need a second server as a reserve. The contracting party is our hardware partner FAST LTA. The service contract is mandatory when buying the server, costs 9% of the purchase price per year and runs for 5 years.
Our hardware partner FAST LTA builds and sells the servers and the service contract. Three weeks after the order is received, the server is with you, installed in your data centre, with an introduction to the server. basebox sets up the platform and provides the licence, support and services.
Yes. basebox runs on your hardware in your network, with sign-in, permissions, apps, connectors and log. The AI models come from another location, for example from your data centre or from a provider you connect. With a third-party provider, we point out the consequences for data protection. You will find the price in your personal quote.
Hospital & clinic
To any system that provides its data through an interface. For the HIS (hospital information system), the vendor grants this access, for example via the FHIR interface under ISiK, which German law provides for hospital systems, via a data platform of your hospital, for example based on openEHR, or via an interface your hospital already uses. Once access is in place, we build the connector, coordinated with you and the vendor. The same applies to laboratory, radiology, billing and nursing documentation. Many use cases run without these systems, using documents, guidelines and e-mails. All connectors →
No. basebox works alongside the HIS. The AI in the HIS covers the HIS; basebox covers the whole hospital: documents, guidelines, e-mails and other systems. If the HIS is connected, basebox also uses its data.
Hospitals and university hospitals, such as Klinikum Nürnberg, Universitätsklinikum Leipzig, TUM Universitätsklinikum, Gesundheit Nordhessen and Klinikum Itzehoe. At the University of Duisburg-Essen, the team of Univ.-Prof. Dr. Keno Bressem analyses patient data with basebox on a server on its own premises.
Yes. Your environment in the German data centre is separated from other customers, the data is never used for training, and the DPA is part of the contract. This applies to the server anyway: the server belongs to you. C5 and Section 393 SGB V →
In the cloud €47,932 per year, around €240 per person. With a server, €109,643 once for the server, plus €21,569 licence and €9,868 service contract per year; anyone may sign in. Non-profit organisations receive 15% off the licence. List prices net. Calculate for your number → What AI costs in a hospital →
Local government & authorities
For example residents' registration, document management, council information system or budget. In the quote, we check each connection individually. Connectors for public administration →
There is a dedicated package for them with answers on logging, permissions and co-determination, plus a model staff agreement. For works and staff councils → Template and process →
Yes. The cloud starts from 5 people at €490 per person per year; you pay only for the people who use the AI. Calculate the price →
Yes, on the server and in the cloud. The data stays in the EU, hosting is in Germany, processing is assigned to a person and role, and the DPA under Art. 28 GDPR is part of the contract. Your data protection officer carries out the case-by-case assessment with our documents. Data protection →
In the cloud €47,932 per year; non-profit organisations receive 15% off the licence. With a server, €109,643 once for the server, plus €21,569 licence and €9,868 service contract per year. List prices net. Calculate for your number →
Yes. basebox also runs at your data centre or municipal IT association; the licence then applies to its server. Talk to us and we will coordinate with the service provider. Contact →