Many small teams working with geospatial data—municipal planners, field
surveyors, local governments, and non-profit organizations—face a common
problem: they collect and maintain geographic data internally, but lack the
engineering capacity to build and operate their own spatial data infrastructure.
Existing solutions either require significant server administration expertise
or push data into external cloud services where teams lose direct control.

geo-base is an open-source, self-hostable geospatial data platform designed
for exactly this context. It provides a tile server, a web-based management
dashboard, and a built-in Model Context Protocol (MCP) server as a single
deployable system. Teams can upload raster and vector data in open formats
(GeoTIFF, PMTiles, GeoJSON), preview and manage tilesets through a browser
interface, and query their data using natural language via MCP-compatible AI
clients—without sending data to third-party services.

The entire stack is built on open-source components: a tile server API,
a web-based management dashboard, an MCP connector layer, a spatial database,
and a browser-based map renderer. The system is designed to be deployable by
a single person with basic command-line familiarity, without dedicated
infrastructure or database administration staff.

This talk shares the motivation behind geo-base, the design decisions that
prioritize operational simplicity over feature breadth, and honest reflections
on where the approach works well and where it falls short. The session includes
a live demonstration showing how a non-technical team member can upload a
dataset and immediately query it in natural language through an AI agent.
We hope this sparks discussion on how the open-source geospatial community
can better serve small, non-IT organizations as primary users.