An Intelligence Briefing for the Port of Rotterdam, from a Single Prompt

An MCP-enabled language model was tested on live maritime data to produce an intelligence briefing for the Port of Rotterdam.

It queried VesselAPI and related datasets to pull recent vessel movements, ship details, EU MRV emissions records, port state inspections, and NAVTEX warnings.

The model used parallel tool calls where possible and then assembled the results into a structured report.

The experiment shows that grounding LLMs in external maritime databases can reduce hallucinations and improve trustworthiness for automated reporting.