Hermes

Introduction

What Hermes for GIS is, and how the configuration is put together.

Hermes for GIS is the documented configuration behind Mercator, a working GIS agent that runs on the open-source Hermes agent from Nous Research. It is a summary of a real setup: the toolsets, MCP servers, the local GIS stack, and a library of GIS skills you can pull from GitHub into your own Hermes profile.

What is what

Hermes is the agent, a self-improving CLI/TUI agent with a learning loop, skills, and a toolset system, built by Nous Research (MIT). Hermes for GIS is this configuration: the tools, stack, and skills that turn Hermes into a GIS agent. It is an independent project and is not affiliated with Nous Research.

What the configuration includes

  • A GIS toolset, exposing esri, open-data, and postgis tools the agent can call through its normal <tool_call> function calling.
  • MCP servers for Esri, Census, open data, and research.
  • A local GIS stack on the machine: PostGIS, QGIS, GDAL, DuckDB, Tippecanoe, and a Python geo environment.
  • A skill library, 21 agentskills.io skills in ~/.hermes/skills/gis/.

What Hermes brings

  • A model-agnostic runtime that runs on the provider you configure per profile.
  • A learning loop that curates and refines skills across sessions.
  • Cross-session recall, a full TUI, and messaging gateways.

Where to start

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
git clone github.com/daraobeirnecode/hermes-gis-skills ~/.hermes/skills/gis
hermes

Then read Installation for the full setup, or Configuration summary for the complete inventory of toolsets, MCP servers, stack, and skills.

On this page