
Founded the agency IT-WINGS in 2016 and still runs it. It serves well-known clients in the automotive, consumer goods, pharma and mechanical engineering industries.
We built Propspector and we run it ourselves. The software reads the documents on an apartment or house against a fixed review catalog, from the land register to the building file. The report shows every deviation with its source reference. It also includes a market value estimate via an external API.
Propspector checks each detail on its own, then each one against all the others. Click through an example (the demo is in German).
Das Exposé gibt 48,12 m² an, die Wohnflächenberechnung weist 41,55 m² nach WoFlV aus. Eine Abweichung von 15,8 %.
Dem Grundriss fehlen Maße, ein Maßstab und die eindeutige Objektzuordnung. Das Dokument ist in dieser Form nicht für eine Finanzierungsprüfung geeignet.
Im Grundbuchauszug sind Nachträge vom 12.02.1992 und 19.03.1992 vermerkt. Diese fehlen in den vorliegenden Unterlagen der Teilungserklärung.
Documents arrive unsorted and badly named. Propspector identifies the documents, assigns them to the right part of the report, cross-checks them and flags what is missing.
Propspector reads the minutes of every meeting and every circular resolution, tracks topics across the years and collects all approved special levies in one place. It also catches a missing year of minutes.
Findings are sorted into “clarify before purchase,” note and no issues. Each one names the document and the page it comes from. The report runs in the browser and on a phone and also comes as a PDF.

A single AI pass misses things or invents them. That’s why Propspector works as a chain of specialized reviewers. The language model pulls the printed facts out of the documents. Hard-coded logic handles the calculating, comparing and classifying.
Each document type has its own reviewer with a fixed catalog. The model extracts the facts, each with its source reference.
Appraisal firms and advisors pass the report on to their clients under their own name. So we built an editor right into the report.

The “Customize PDF” editor with sample data. Click to start the video.
Third-party software creates the property and uploads the documents. The finished report comes back as structured data and as a PDF. AI assistants access it directly via MCP. We currently offer the API and MCP as a pilot, on request.
Three points. According to the building permit, the attic doesn’t count as living space, the energy source differs from the building file, and the plot size doesn’t match the land register.
Eight tools. Starting a review is marked as a modifying action. Before it uses credit, the assistant checks with you.
We have been working on Propspector for almost two years. We developed the review catalog with appraisers, financing advisors and real estate dealers. Every case that shows something new helps us refine it.
Anyone who processes other people’s purchase documents is responsible for their data. That’s why data protection is built into our architecture.

Founded the agency IT-WINGS in 2016 and still runs it. It serves well-known clients in the automotive, consumer goods, pharma and mechanical engineering industries.

Has been scaling digital businesses since 2015, first at Festool, then for six years at Mercedes-Benz. There he led projects in digital products and global performance marketing.
Propspector is a software product by Kessel & Code. Using AI, it reads the purchase documents for apartments and houses against a fixed review catalog and produces a report that lists every deviation with its source reference.
For anyone who reviews real estate files, such as appraisers, mortgage advisors and investors. Private buyers also use it before buying. Details and pricing are on propspector.de.
No. The report isn’t a market value appraisal or legal advice. Propspector does the reading and comparing. The on-site assessment and the professional decision stay with people.
A chat answers freely. Propspector follows a fixed review catalog. The model extracts the facts from the documents, and deterministic code handles the comparisons and classifications. Every finding points to its source.
Yes. The application and documents sit on servers in Germany, and the AI analysis runs in an EU region. Business customers get a data processing agreement under Art. 28 GDPR.
MCP (Model Context Protocol) is an open standard that lets AI assistants such as Claude and ChatGPT access software. With the Propspector MCP server, you ask your assistant about properties and reports. Before a review uses credit, the assistant checks with you.
Yes, as a pilot, on request. Your software creates a property, uploads the documents and gets the report back as JSON and PDF. It’s the same review as in the app.
Yes. The same approach works for contracts, minutes or technical documents. We then build the review catalog with your specialists.
We use the same approach for your documents and processes. Tell us briefly what you need. You’ll get an answer within 24 hours.
hello@kessel-und-code.de
Rotebühlplatz 11–15, 70178 Stuttgart