OpenStreetMap (OSM) is the most comprehensive open-source geospatial database in the world—but as it continues to grow, the tools for understanding and managing this data need to evolve as well. In this workshop, we explore how AI, when thoughtfully integrated with OSM workflows, can offer intelligent, context-aware support to mappers and developers alike.

This session introduces participants to the Model Context Protocol (MCP)—a lightweight, structured format for embedding task-specific context into AI interactions. MCP allows for the creation of prompts and workflows that are reproducible, local-first, and aware of the mapping domain. Using this protocol, we will build a proof-of-concept tool that uses open-source large language models (LLMs) to interpret OSM data and assist with mapping tasks, such as tag validation, feature classification, and geospatial query answering.

Importantly, the workshop focuses on privacy-respecting, open-source solutions. All AI tools used will run locally on participant machines, avoiding any reliance on commercial APIs or user-tracking platforms. This aligns with the ethos of the OSM and open-data communities, ensuring participants can experiment with powerful AI without compromising control or transparency.

Attendees will gain practical skills in working with Overpass API to extract OSM data, creating MCP-compliant prompts, running local LLMs via tools like LM Studio or Ollama, and designing AI agents that enhance mapping workflows. No prior experience with AI is required—just a curiosity for what happens when maps meet models.