We set up a working environment in which you extend your own IT with a coding agent such as Claude Code: automating workflows, connecting systems, producing documents and analyses. This includes the technical infrastructure, a context system, security rules and your enablement. The first solutions are built during the project, all further ones you build yourself.

A coding agent is the combination of an AI model and a program that gives this model tools: reading and writing files, running commands, calling interfaces. The two are matched to each other. Such an agent writes software, and beyond that it uses existing programs, your computer and any kind of interface (API, CLI, MCP). You describe in plain language what you need; the agent implements it, tests it and connects it to your systems. We work with Claude Code; a tested, comparable open-source alternative is OpenCode.

Yes. The agent writes the code. Your job is to know what your business needs and to decide: what gets built, how it should behave, whether the result is right. What you need to bring is the willingness to engage with the technology. Our reference clients are an estate agent and a structural drying company.

Chat assistants answer questions and write texts; they work inside one application or are (web) applications themselves. A coding agent works across your entire system landscape: it creates records, generates documents, connects programs and builds functions that run permanently. For texts and research, your AI platform remains the right tool. We start where it stops.

No. Vibe coding means telling an agent roughly what you want and running the result unchecked. That is legitimate for a prototype, but not enough for day-to-day operation. With us, the agent works within a fixed framework of infrastructure, rules and context. Everything is versioned and documented, and you understand what has been built.

We build with you. In addition to a coding setup and your general enablement, we work out optimisations and automations together. AI allows us to use digital technology in an entirely new way and at a new speed. Together we build a technical foundation and key functions; after that, you are able to develop and extend your digital infrastructure yourself.

In principle, almost anything. Common use cases are: voice agents for data entry by dictation directly into your business software, automated workflows for enquiries and correspondence, and documents in the format that clients, insurers or authorities require. Lead generation and sales funnels can also be implemented very well and connected to existing CRM and ERP solutions. In teams, data pipelines, analyses and reporting are added.

As an owner-managed business, a business software with an API, that is, an interface through which other programs can read and write data. Suitable examples are onOffice with the API module, Propstack, Lexware Office from XL upwards, Odoo, casavi, HubSpot and Pipedrive. Excel and desktop software without an API cannot carry the project; we support a change of system on request. Teams need Git hosting, a cloud project or a server under their own administration, and Claude accounts with team administration.

In four steps. In the free initial call we clarify your requirements and your situation; afterwards you receive an offer with a project plan. Then we set up the infrastructure together and derive an implementation concept from your processes. In the build phase, the agreed functions are built together, with a growing share on your side. After the handover, we are available for questions for two months.

That also depends on how much time you can free up for the joint work. A project duration of about four to six weeks is usual and manageable for most.

We work for a fixed price; price indications can be found in the details that you can download on this page. In addition, there are running costs for the services used: the Claude subscription, a small cloud infrastructure and, depending on the provider, the API module of your business software. There are no licences from us and no maintenance contract as a condition.

The system belongs to you and runs in your accounts; we keep no access. Everything is documented in such a way that you and your Claude Code instance know all the details. You make extensions and corrections yourself.

Our projects work very well fully remotely. In-person meetings within Germany, Austria and Switzerland are possible by arrangement.

Business software, cloud project and repositories run in your accounts and in your name. What the agent works on is sent to Anthropic, the provider of Claude, for processing; in the business plans, this content is not used to train the models. In an open-source setup (OpenCode), you are free to choose the model provider. We advise on all elements of the system with regard to the GDPR and the EU AI Act. The responsibility for what you implement remains with you, however.

The agent needs access to your systems, and this access is the most sensitive part of the whole setup. Passwords and API keys are kept in a password vault and never in files, in the code or in the repository; programs fetch them from there only at runtime. We set up this secrets management together with you and establish the rules the agent follows.

Your business software remains the core and handles business logic and bookkeeping as before. What you build is an extension next to it, with limited permissions. Every change is versioned and can be reverted, and during the project we review everything you build thoroughly and also establish test mechanisms.

No. The extension uses the official interface that the provider offers for this purpose. Nothing is changed in the software itself.

Hardly. What is built consists of open standards: scripts, a database, repositories and a context system made of simple text files. All of this can be continued with another coding agent such as OpenCode.

The platform covers the breadth: texts, summaries, search in documents. Our approach is aimed at the few teams whose tasks go beyond that, for example analyses across several data sources or concepts that need the state of three weeks and four heads. Both can exist side by side.

It sets up permissions, cost limits and secrets management with us and keeps control of the environment. Within this framework, the team works independently.

For anyone looking for an AI that does everything without their own involvement. Our offer requires that you or your team want to take your own digital infrastructure into your own hands.