Geo

claude mcp add --scope project clio-geo -- clio-kit mcp-server geo

Install the launcher first. Restart your client after setup. Configuration locations →

clio-kit doctor --server geo --connect

Tools​

render_feature_map

Render one or more GeoJSON layers (polygons/lines/points) onto a single map PNG with an optional basemap. Each layer accepts a style with fixed colors or data-driven coloring (color_by + category_colors, an 'epa_aqi' AQI scale, or a matplotlib colormap). Returns the output path and bounds.

points_in_polygons

Spatial overlap: return which GeoJSON points fall within (optionally buffered) GeoJSON polygons — e.g. which AirNow monitors lie inside the smoke footprint. Accepts inline GeoJSON or file paths. Returns the matched points with their properties and a matched_count.

bounding_box

Compute the bounding box [min_lon, min_lat, max_lon, max_lat] of the VALID geometry in GeoJSON features (inline or file path), optionally padded by pad_km and rounded to 4 decimal places. Features whose geometry cannot be parsed are skipped, so feature_count reports valid geometries only. Use this to derive an analysis region for mapping or a spatial query; use feature_bbox instead to measure a document's raw coordinate extent including malformed features.

query_arcgis_features

Query an ArcGIS FeatureServer layer (with optional lon/lat bbox and where clause) and write the returned features to a local GeoJSON file. The saved FeatureCollection can be fed directly into render_feature_map, points_in_polygons, or bounding_box.

geocode

Look up a free-text place name or location and return real coordinates from OpenStreetMap Nominatim (a lookup, not a model guess). Each match carries lat/lon, a [min_lon, min_lat, max_lon, max_lat] bbox, type, importance, and a provenance source so the region can be grounded and cited.

filter_points_by_radius

Filter/rank any table of points by great-circle distance to a center. Reads a CSV (or GeoJSON points) of locations, computes the haversine distance from (center_lat, center_lon) to each row, and returns the rows within radius_km sorted ascending by distance, each annotated with distance_km. Latitude/longitude columns are auto-detected from common names (latitude/lat, longitude/lon/long) when not given. Domain-neutral: works for sensors, sites, cities, samples, or any lon/lat table.

inspect_geojson

Inspect a GeoJSON document and report its geometry types and counts, feature count, property keys (schema), bounding box [min_lon, min_lat, max_lon, max_lat], CRS if present, and total vertex count. Reads the document as written, without a geometry engine, so it reports what the file actually contains. Accepts a file path or inline GeoJSON.

validate_geojson

Validate the structural well-formedness of a GeoJSON document: that the top-level type is recognized and every geometry's type and coordinates are well-formed (correct nesting depth, finite numeric positions). Returns {valid, errors}. Run this before a rendering or overlap tool, which silently skip geometry they cannot parse.

summarize_geojson

Produce a compact human-readable summary of a GeoJSON document: counts per geometry type, bounding box, property keys, and a few sample feature property sets. Accepts a file path or inline GeoJSON.

feature_bbox

Compute the bounding box [min_lon, min_lat, max_lon, max_lat] of EVERY coordinate in a GeoJSON document, without validating geometry and without rounding. feature_count counts all features in the document, including malformed ones. Use this to inspect a file's raw extent; use bounding_box instead for the extent of valid geometry, padded and rounded, when feeding a map render or spatial query.

Examples​

Replace example paths with your own files and check the returned results.

Validate and map geospatial data​

Start with validate_geojson, then inspect extent with feature_bbox. Confirm the data's coordinate reference system before testing spatial relations with points_in_polygons. Use render_feature_map to render the selected layers to a PNG and verify the returned file.

Web basemaps and remote feature queries need network access. Use local geometry for a reproducible check, and distinguish a missing basemap from invalid input. The server guide describes layer styling and output fields.