Claude consultant for companies in Germany
I am Michael Wutzke, an IT expert for data platforms in Frankfurt, and I work with Claude daily in a client's code and data. I set Claude up for your developers and write the rules for the code it generates. Tasks in your repository I take over myself. I support companies as a freelancer or as a permanent employee, remote for the most part with on-site days in Frankfurt Rhine-Main. Book a 30-minute video call. We look at your task, and you find out whether I can help.
What your company gets
-
Experience with data platforms
For more than twenty years I have built software and the data platforms behind it. Claude speeds up that work, and I know which generated change can damage your records.
-
Claude in client work
For a client I break up a monolithic platform with Claude and Codex. The project has more than 300 RESTful API endpoints and more than 8,000 commits.
-
Rules in your repository
The rules for Claude sit in one file in your repository, and hooks check the hard ones before a call runs. Your developers read and change them in the same place.
-
Review before the merge
Generated code reaches your main branch through a merge request, as a colleague's code does. A read-only script checks what an agent reports about its own work.
-
Your data under your control
Claude starts with read rights on a development database. A write to live data waits for a person's confirmation, given outside the chat.
-
Teaching on your code
I teach practical work with Claude at Claude Hacker House. Your developers learn the setup on your own code while we work together.
What a Claude consultant does in your company
A Claude consultant brings Claude into the code and data work of a company and does part of that work with it. When each developer runs Claude with personal settings, nobody has written down what the tool may change. I set it up once for the team and write the rules. I also take on tasks in your repository and your database, with Claude as my tool.
Where Claude saves time, and what stays with a person
Claude and Codex write the code and the drafts. The data model, the review and each production change stay with a person. The table shows how I split the work on a client’s information platform and on this site. You can hold your own backlog against it before our call.
| Task | What the agent does | What stays with a person or a check |
|---|---|---|
| New API endpoints on a monolithic platform | Writes the code, with Claude and Codex in one repository | The data model and the review of each result |
| A change in code the agent has not read before | Queries a knowledge graph of the repository for the callers and earlier decisions first | The review of the merge request |
| Moving a website to a new system | Did much of the migration work when this site left WordPress | The choice of the target system and the layout check at phone and desktop width |
| Duplicate records in a database | Checks the candidate pairs that fixed rules cannot settle; “unsure” leaves both records in place | Where automatic merging stops, and a checked sample in each report |
| Longer texts such as documentation | Writes the draft, and a hook refuses it until the agent has reviewed it | Judging the text, which the hook cannot do |
| A report on a finished job | Reports what it changed | A read-only script counts the result; its number stands over the agent’s report |
| A write to live data or a push to the production branch | Nothing: a hook blocks the call | A person confirms each production change outside the chat |
| A new type in the data model or a question of classification | Nothing | A person decides |
How I set Claude up in your team
I start with the rights, write the rules next and add the checks last.
First the rights. Claude reads the development database and cannot write to it, and its credentials cannot push to the production branch.
Then the rules. One file in the repository holds them, one line per hard rule, and each line points to the document behind it. Longer instructions become skills, which Claude loads when a task matches.
Then the checks. A rule whose break would do large damage becomes a hook that blocks the call before it runs. Each hook gets tests for the calls it must allow, because developers switch off a check that blocks valid work. A team that needs more machine checks than these gets a full agent harness for its repository.
Rules and review for generated code
Generated code goes through the same review as code from a colleague. On the client platform each code agent works in its own git worktree, commits under its own name and opens a merge request. A push waits until the local copy of the CI checks has passed. The review of each result is my part of the work, together with the data model and the workflow behind the software.
Codex next to Claude
Codex is OpenAI’s coding agent. On the client platform it works in the same repository as Claude, under the same rules. The two read their instructions from different files. One file is the source and the other imports it, so a rule is written once and holds for both. A team that works with Codex alone keeps the rights and the rule file of this setup, and its hooks are written for Codex.
Who I am
I am Michael Wutzke, an IT expert for data platforms in Frankfurt. My path runs over more than twenty years from code to CIO, in finance, e-commerce and real estate, with cloud platforms and data centers. Today AI agents speed up how I build and transform data platforms. I teach practical work with Claude at Claude Hacker House, where my own sessions cover MCP servers and virtual organizations of agents. I hold a training authorization from IHK Frankfurt for two IT apprenticeships.
How we start
-
Video call
In 30 minutes we talk about your code base, your data and the tasks your team wants to hand to Claude.
-
Current setup
We go through the accounts, rights and personal settings your developers use today, and pick the task Claude takes over first.
-
Offer or quote
For a permanent position you send an offer. For a freelance project I send a quote. The work starts once both sides have signed.
-
First task
I take one task from your backlog and do it with Claude under the new rules, in your repository, with your developers in the review.
-
Handover
At the end of a freelance project your developers own the rule file and the hooks. Each rule points to the document that explains it.
Questions companies ask
Which engagements do you take on?
Permanent positions, remote for the most part with on-site days in Frankfurt Rhine-Main or at your office. Freelance projects, part time or full time.
Does Claude also work on data, or only on code?
On data as well. On a client's information platform, AI agents check duplicate candidates that fixed rules cannot settle, and an unsure answer leaves both records in place. Questions of classification and new types in the data model stay with a person.
Our developers already use Codex. Does that change anything?
Codex stays. In client work Claude and Codex read their rules from one source file, so a rule is written once and holds for both. Hooks are written for the tool that loads them, so each tool gets its own.
Which data does Claude see?
Your company decides. Claude starts with read rights in the development environment, and a production change needs a person's confirmation. Data that may not go to a cloud model can stay with an open-weight model on your own server.
Do you also train the whole team?
Yes. During a project your developers learn on the tasks we do together. For staff outside development I run a Claude training, online or on site.
What does a company need before you start?
Claude accounts under your company's plan, access to a development environment and a developer who reviews merge requests. We check these points in the first call.
Details on the work behind this page
-
AI agents under the same rules as humans
Limits at the resource, a blocking check before each tool call, and approvals that only a person gives.
-
Building software
I use AI to build MVPs and production-ready software quickly, above all for data management and workflows.
-
Change management and workflows
Merging duplicates and changing data at scale can destroy good records. I work on changes that can be undone.
-
Teaching and certifications
Where I teach, my certificates with the issuer of each, and my education.
Your Claude consultant in Germany
I am Michael Wutzke, a Claude consultant in Germany, based in Frankfurt. In a 30-minute video call we talk about the code and data your team works on. You learn which tasks Claude can take over there and which stay with your people.
Searches this page answers
- Claude consultant
- Claude expert
- Claude AI specialist
- Claude freelancer
- Claude best practices for coding
- Claude for software development
- Claude for coding
- Claude for data engineering
- Claude for documentation
- Claude and Codex
- Claude and Codex working together
- OpenAI Codex for teams
- AI freelancer
- AI coding consultant
- AI assisted software development