# BelegBot: AI-assisted receipt processing as our own product

> An in-house product for automated receipt processing: submit receipts, extract data via LLM, validate and export — including an MCP interface.

Quelle: https://kontrollfeld.de/en/case-studies/belegbot/
Sprache: en
Kunde: Kontrollfeld (own product)
Branche: Product Development
Tech-Stack: Next.js, Prisma, LLM-APIs, MCP, Docker
Veröffentlicht: 2026-02-20
Hinweis: KI-generiert, menschlich reviewt.

---
## Starting point

Collecting receipts, typing them up, assigning them: in many organisations, receipt processing regularly consumes time better spent elsewhere. We wanted to know how far the process can be reliably automated with current LLMs — and turned the answer into a product.

## Solution

BelegBot accepts receipts as photos or files, extracts the relevant fields via LLM, validates the results against rules and provides the data for accounting and export. An MCP interface additionally exposes its capabilities to AI assistants. It runs as a container on our own infrastructure.

## Outcome

A production system we use ourselves — and practical proof that we take LLM-assisted automation from idea to operated product.