Hermes for GIS: a GIS configuration for Hermes

This is a summary of the Mercator configuration: what is enabled, what it connects to, and what runs locally.

Hermes profile

Setting	Value
Profile	`mercator`
Config	`~/.hermes/profiles/mercator/config.yaml`
Terminal backend	local shell
Context compression	enabled
Memory, session search	enabled

The model is chosen per profile; Hermes is model-agnostic and runs on the provider you configure.

Enabled toolsets

web  browser  terminal  file  code_execution  vision
image_gen  tts  skills  todo  memory  session_search
clarify  delegation  cronjob  computer_use

MCP servers

Server	Role
`esri`	ArcGIS Online / REST search, layer inspection, queries, geocoding
`census`	Census API access and demographic enrichment
`open-data`	Public data portal search and fetch
`composio`	Connected apps (GitHub, Sheets, Drive, and more)
`notebooklm`, `exa`	Source-backed and deep research
`orgo`, `hetzner`	Sandboxed execution and infrastructure

Local GIS stack

~/.hermes/.env                 credentials (ARCGIS_TOKEN, POSTGIS_URL)
gis_lab (PostgreSQL 18.4)      PostGIS 3.6.3
  schemas: source, staging, analysis, app_public, audit
QGIS 4.0.3                     desktop and processing
GDAL / OGR, Tippecanoe, DuckDB format, tiles, analytics
gis-lab/.venv                  GeoPandas, Shapely, Rasterio, pyproj, Fiona, OSMnx
gis-lab/.venv-esri             ArcGIS API for Python

Skills

21 GIS skills in ~/.hermes/skills/gis/, pulled from github.com/daraobeirnecode/hermes-gis-skills. See Skills for the list.

Environment variables

Variable	Used by
`ARCGIS_TOKEN`	`esri` tool, ArcGIS API for Python
`POSTGIS_URL`	`postgis` tool

Open data portals and Census are public, so those tools need no credentials.

Configuration summary