Swiss Grounding MCP

At the Swiss {ai} Weeks Zurich Hackathon on 24–25 September 2026, I joined Swisscom’s Swiss Grounding MCP challenge. Our international team of seven met at the event, came together at 13:00 on 24 September, and worked until noon the following day. The team comprised Roberto Cerrone, Edoardo Diana, Alberto Minetti, Vincent Van Loo, Victor Bonilla, Jesus Sebastian, and Jiaqi Yu.

The challenge was to build a Model Context Protocol (MCP) server that helps AI assistants answer questions about Switzerland using authoritative public sources, with attention to the relevant jurisdiction and verifiable references. Together, we developed a Python server using FastMCP and SQLite. The team’s implementation brings together tools and data for Swiss public information, including the commercial register, health insurance, housing, and driving licences.

My contribution focused on housing and the exchange of foreign driving licences. I developed three MCP tools for the mortgage reference interest rate, housing information, and driving-licence exchange, backed by data imported into SQLite. I also worked on building the local databases when the server starts. These changes are documented in my merged pull request. I tested the team’s MCP server with OpenCode and Claude, running multiple agents in parallel to compare how the models behaved with and without the server. This work was part of the team’s shared submission.

The submission, listed as dantevier in the Swisscom challenge evaluation, placed second among six teams by event score, with 3.30 out of 5. The separately published quality ranking placed it fourth among the four ranked teams.

Zurich Hackathon · Swisscom challenge · Challenge documentation · Team repository