MOHRWURM(1)Martin Mohr · Software & Web

NAME

Automation, integrations & AI

If someone on your team copies data between systems by hand every week, that can usually be automated. I use AI where it takes work off your plate, and leave it out where it does not.

TYPICAL WORK

Connecting systems
Shop, inventory, CRM or accounting exchange data directly through their APIs instead of export and import.
Recurring workflows
Reports, reconciliations and notifications run on a schedule and speak up when something is off.
AI for text and documents
Pre-sorting emails, reading documents, writing summaries or drafts with models such as Claude or GPT. Where it matters, a person checks before anything goes out.
MCP interface
Through the Model Context Protocol, AI assistants work with your data directly, for example: "Which orders are still open this week?" Read-only or with clearly defined actions.

PROCESS

  1. ConversationYou tell me what it is about. I ask questions and look at what is already there: workflows, data, existing code.
  2. ProposalI write down what I would build, in which steps and with how much effort. Then you decide.
  3. Building in stepsYou see a working version early. That way we notice quickly if something should be different.
  4. Handover or operationsYou get the code, the docs and all access. If you want, I keep running and maintaining the application.

STACK

python · typescript/node · php · rest apis · claude & openai api · mcp

EXAMPLE

market-screener

Several times a day the screener pulls market data from different sources, checks it against fixed rules and sends hits via Telegram. Over MCP it can be queried straight from an AI chat. packeroo and wickwatch have an MCP interface too, which I use to ask for figures and status in a chat.

QUESTIONS

What is MCP?
The Model Context Protocol is an open standard that lets AI assistants such as Claude use tools and data. Instead of pasting data into the chat, the assistant fetches it itself, and only what the interface allows.
What happens to my data at an AI provider?
I show you up front which data goes where. I use the API access of the providers, where inputs are not used for training, and send only what the task needs. Some things can be solved without AI altogether.
Is this worth it for a small company?
Often especially there. Two hours of reconciling per week add up to about a hundred hours a year. Whether automating it pays off is something we estimate together in the first conversation.
What if an API changes?
Automations often fail without anyone noticing. That is why I add monitoring that speaks up when a run fails or does not happen.

SEE ALSO

Tell me briefly what it is about. I usually reply within two working days.