GeoAgent

GeoAgent is a QGIS plugin for geospatial analysis in plain language. Ask it to load data, explore your layers, or run a chain of processing steps: it picks the QGIS algorithms, fills in their parameters, runs them, and adds the results to your map.

New to GeoAgent? Install the plugin, connect an LLM provider, then follow the quickstart.

What it can do

• Explore and manage data — add layers from files or URLs, list layers and columns, zoom, select features by attribute or geometry, and create, load or save projects (General mode).

• Run geoprocessing in plain language — any algorithm in the QGIS Processing Toolbox (native, GDAL, GRASS, …), alone or as a multi-step chain where each result feeds the next (Processing mode).

• Turn a run into a QGIS model — open what ran in the Model Designer, or save it as a .model3 file to re-run on other data (Model export).

• Use the LLM you prefer — Ollama (local and free), OpenAI, Google Gemini, or Anthropic Claude (LLM providers).

• See what happened — every step, the parameters used, and the tokens each request cost appear in the Logs tab (Logs and token usage).

License

GeoAgent’s code is released under the MIT License. When it runs inside QGIS (GPL v2+, with PyQt under GPL v3), the combined work is governed by those licenses’ terms, as with any QGIS plugin.

Citation

If you use GeoAgent in your research, please cite the preprint:

Kshetri, T. B., & Ojha, R. (2026). GeoAgent: An Agent-Based QGIS Plugin for Natural-Language-Driven Geospatial Analysis with Multi-Backend Large Language Model Support. EarthArXiv. https://doi.org/10.31223/X53803

@misc{kshetri2026geoagent,
  title     = {GeoAgent: An Agent-Based QGIS Plugin for Natural-Language-Driven
               Geospatial Analysis with Multi-Backend Large Language Model Support},
  author    = {Kshetri, Tek Bahadur and Ojha, Rabin},
  year      = {2026},
  publisher = {EarthArXiv},
  doi       = {10.31223/X53803},
  url       = {https://doi.org/10.31223/X53803}
}