GeoEDGE MCP

Clip a raster dataset using polygons from a shapefile and write the result. Converts the shapefile's CRS to match the raster's CRS if they are different. The output is a rectangle, so some of it necessarily falls outside your polygon(s). That margin is written as the source's declared nodata value. When the source declares none, the result is promoted to float and declares NaN as its nodata rather than filling the margin with 0 — a 0 there would be indistinguishable from a measured 0 and would pull down every mean, minimum and histogram computed on the clipped raster afterwards. Parameters: …

Clip vector geometries using geopandas.clip. Args: gdf_path: Path to the input geospatial file. clip_path: Path to the clipping geometry file. output_path: Optional path to save the result. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status, message, and output info.

Close the map editor and stop its local web server, releasing the port. Returns: Dict confirming whether the editor had been running.

Compute NDVI (Normalized Difference Vegetation Index) and save to GeoTIFF. Pixels the source declares as nodata (a fill border, a scene edge, a masked cloud) are excluded from the maths and written as nodata in the output, and the output declares NaN as its nodata value rather than inheriting the source's. Both matter: a Landsat Collection 2 scene fills with 0, and an NDVI raster that inherited `nodata = 0` would discard every genuinely bare, urban or rocky pixel (NDVI near 0) as if it were missing data. Parameters: - source: input raster path. - red_band_index: index of red ba…

Compute NDVI from Sentinel-2 L2A (B08=NIR, B04=Red) for an AOI/time window via MPC. - Picks a low-cloud item intersecting the AOI. - Clips to AOI and writes single-band NDVI GeoTIFF scaled to float32 (-1..1). Args: datetime: STAC datetime/interval (e.g., "2024-07-01/2024-07-15"). cloud_cover_lt: Only consider items with eo:cloud_cover < value. Use None to disable. bbox: "minx,miny,maxx,maxy" (WGS84). geometry_geojson: GeoJSON geometry (WGS84). out_crs: Output CRS for NDVI raster. filename: Output filename (default auto-generated). path: Output directory. Returns: …

Concatenate multiple single-band raster files into one multi-band raster, handling alignment issues automatically. Parameters: - folder_path: Path to folder containing input raster files (e.g. GeoTIFFs). - destination: Path to output multi-band raster file. - overwrite: When False, refuse to run if destination already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. - resampling: How to resample a band that has to be warped onto the reference grid: "nearest" (default -- never…

Calculate convex hull of a geometry.

Create a styled map from multiple inputs (vectors, rasters, WKT, or coords). Args: layers: List of dicts, each containing "data" and "style". data can be: any OGR-readable vector file path (.shp/.geojson/ .gpkg/...), a raster file path (.tif/.tiff/.img/.jp2), a WKT string, coords, a GeoDataFrame, or a name registered via register_layer (resolved automatically). "style" for VECTOR layers may include: {"color": "blue"} # plain solid color {"column": "type"} # categorical/unique-value coloring {"column": "value", "…

Render a print-ready suitability map (PDF/PNG) from a suitability raster. Produces a cartographic product with a graduated colorbar legend, optional scale bar (projected CRS only), north arrow, and title block + attribution footer. Optionally overlays ranked candidate sites and a context boundary. Branding defaults (colour ramp, attribution) come from the named map ``template`` unless explicitly overridden by ``cmap`` / ``attribution``. Args: suitability_raster: Path to the 0-1 suitability raster (band 1). destination: Output path; extension sets the format (.pdf or .png). title…

EXPORT a standalone interactive map as a .html FILE, to send to somebody else, attach to a report, or keep. Call this only when the user asks to export, share, save, send or email a map, or explicitly asks for an HTML file. To SHOW the user their data, use ``visualize_data`` instead — it puts the data straight onto an editable map in their browser and needs no file. This tool writes a file to disk; that is its purpose, and it is the wrong answer to "show me". Supports both vector and raster layers, and accepts a name registered via ``register_layer`` (or auto-registered by any writing tool) …

Fully probe ONE specific dataset (or one layer inside a multi-layer .gpkg/.gdb) and report its CRS, feature/band count, and extent. This is the expensive, single-dataset counterpart to discover_layers' cheap bulk scan -- call discover_layers first to find what's in a folder and enumerate layer names, then call this once per dataset/layer you actually need details on. Args: path: Path to the dataset (absolute, or relative to storage). For a .gdb, this is the .gdb folder path itself. layer: Optional layer name (from discover_layers' layer_names) for a multi-layer .gpkg/…

Describe a workflow template: its typed parameter schema and plan template. Use this to learn what inputs a workflow needs so you can ask the user only for what is missing before running it. Args: name: Workflow name (see list_workflows). Returns: Dict with status and the workflow schema (parameters + plan template).

Find difference between geometries (``geometry1`` minus ``geometry2``). Args: geometry1: WKT geometry to subtract from. geometry2: WKT geometry to subtract. crs1: Optional CRS of ``geometry1`` (e.g. 'EPSG:4326', 'EPSG:28355'). crs2: Optional CRS of ``geometry2``. When both are given and differ, ``geometry2`` is reprojected onto ``crs1`` first and the result is returned in ``crs1``. Without them the inputs are assumed to share a CRS: a mismatch makes the tool return ``geometry1`` unchanged, which reads as "nothing was clipped" rather than "wrong …

Scan a folder and report every vector/raster dataset found, without reading any dataset's actual features/pixels -- fast enough to point at a whole project folder before doing any analysis. For multi-layer containers (.gpkg, .gdb) every sub-layer name is listed via a cheap fiona.listlayers() call (no full read), so a multi-layer .gpkg is never silently mistaken for a single-layer dataset -- pass one of the reported layer names to describe_layer or register_layer(..., layer=...) to work with a specific layer. A .gdb (File Geodatabase) is a directory, not a single file -- it's always reported …

Dissolve geometries by attribute using geopandas.dissolve. Args: gdf_path: Path to the geospatial file. by: Column to dissolve by (optional). output_path: Optional path to save the result. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status, message, and output info. After dissolving, follow up with create_web_map on the result and relay it in the same response -- don't ask first (see server instructions).

Create a distance-based spatial weights (W) object from point data. - data_path: path to point shapefile or GeoPackage - threshold: distance threshold for neighbours, in METRES. Lat/lon data is reprojected to its UTM zone first, so the number means the same thing whatever CRS the file is in. (A file in a projected CRS whose units are not metres is measured in its own units; the response says which.) - binary: True for binary weights, False for inverse distance weights - id_field: optional attribute name to use as observation IDs

Download GADM administrative boundaries and save as GeoJSON. Args: region: e.g. "USA" or "United States" level: 0=country, 1=state, 2=county, ... path: custom output folder Returns: {"status": "success", "file_path": "..."} or {"status": "error", "message": "..."}

Download climate data (ERA5 or other CDS datasets). Args: variable: e.g. "2m_temperature", "total_precipitation" year: e.g. "2024" month: "01".."12" day: "01".."31" time: Time of day (default: "12:00") dataset: CDS dataset name (default: "reanalysis-era5-single-levels") format: File format ("netcdf" or "grib") path: custom output folder (default: ./data/climate_data) Returns: {"status": "success", "file_path": "..."} or {"status": "error", "message": "..."}

Download analysis-ready satellite imagery from Microsoft Planetary Computer (STAC + SAS). - Picks the least-cloudy item matching your query (by default Sentinel-2 L2A). - Downloads specified asset bands and writes a multi-band GeoTIFF. - Optional bbox/geometry crop and CRS reprojection. Args: collection: STAC collection id (e.g., "sentinel-2-l2a", "landsat-8-c2-l2") assets: list of asset keys to download (e.g., ["B04","B03","B02"]) datetime: STAC datetime/interval (e.g., "2025-08-01/2025-08-31" or "2025-08-05") cloud_cover_lt: only items with eo:cloud_cover < this value. Use N…

Download occurrence records for a given species and save as JSON. Args: scientific_name: Scientific name of the species (e.g., "Puma concolor") limit: Number of occurrence records to fetch (default: 100) path: Custom output folder (default: ./data/ecology_data) Returns: {"status": "success", "file_path": "..."} or {"status": "error", "message": "..."}

Download a street network for a given place using OSMnx. Args: place: Name of the place (e.g., "Los Angeles, California, USA") network_type: Type of network ("drive", "walk", "bike", etc.). Ignored if custom_filter is provided. file_path: Optional. Full path where the GraphML file will be saved. If not set, saves to default location. custom_filter: Optional. OSMnx custom filter string to specify which roads to download (e.g., '["highway"~"motorway|trunk|primary"]'). Returns: NetworkX graph as GraphML file path or error message.

Download ESA WorldCover (10 m) for the given AOI and year from Microsoft Planetary Computer. - Crops to bbox/geometry (WGS84), optional reprojection. - Writes a single-band categorical GeoTIFF (land cover classes). Notes: * On MPC, the collection is commonly 'esa-worldcover' with yearly items; the default here auto-selects by year. If your deployment uses different IDs, pass `collection` explicitly. * `asset_key` is typically 'map'. Args: year: WorldCover year (e.g., 2020, 2021, 2023). collection: STAC collection (default resolves to 'esa-worldcover'). asset_key: Asset key…

Remove a layer from the registry. Does NOT delete the underlying file.

Run dynamic LISA (directional LISA) with giddy.directional.Rose. Returns sector counts, angles, vector lengths, and (optionally) permutation p-values.

Report what has happened in the map editor this session: layers opened, renamed, edited, imported or saved, with the exact layer names to use when asking for further analysis. Also flags any layer with unsaved edits. A layer marked unsaved_changes has been edited on screen but not written back to its source file -- offer to save it before running analysis on it. Returns: Dict with the session id, the current layers and their state, and the change log in order.

Get bounding box of a geometry.

Split multi-part geometries into single parts using geopandas.explode. Args: gdf_path: Path to the geospatial file. output_path: Optional path to save the result. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status, message, and output info.

Extract a specific band from a multi-band raster and save it as a single-band GeoTIFF. Parameters: - source: path or URL of the input raster. - band_index: index of the band to extract (1-based). - destination: path to save the extracted band raster. - overwrite: When False, refuse to run if destination already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior.

Download a spatial dataset from a public http(s) URL into managed storage. Use this for any dataset whose access method is a **direct download** — a zipped shapefile, a GeoPackage, a GeoTIFF, a GeoJSON. The file lands inside the server's storage root and, when it is a format that can be opened directly, is registered as a layer so later tools can refer to it by name instead of by path. A ``.zip`` is opened too: if it holds exactly one spatial dataset, that dataset is registered as a layer — read straight out of the archive where possible, unpacked beside it where the format requires. If the …

Select the features of a vector layer whose attributes satisfy a condition. This is the attribute-query tool: "keep only the polygons that are not Conservation, Residential or Mine land use", "keep parcels over 5 hectares", "drop rows with no zoning code". Filtering happens on attributes only — use ``clip_vector`` / ``sjoin_gpd`` for spatial selection. Refer to columns by name. Comparisons, arithmetic, ``in``/``not in`` against a literal list, and these functions are available: ``isnull notnull isin contains startswith endswith lower upper strip length abs round floor ceil sqrt log exp minim…

Compute focal (moving window) statistics on a raster. Args: raster_path: Path to the input raster. statistic: Statistic to compute ('mean', 'min', 'max', 'std'). size: Window size (odd integer). output_path: Optional path to save the result. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status, message, and output path if saved.

Compute the Gamma index of spatial autocorrelation for a continuous variable. Returns the index (`Gamma`), its permutation p-value, and the mean index under spatial randomness for comparison.

Compute Global Geary's C Autocorrelation Statistic.

Report how GeoEDGE-MCP's one-time setup is going. On a first run the toolset installs in stages so the core vector tools and map rendering are usable while the rest downloads in the background. Call this to see which groups of tools are ready now and which are still installing. No restart is needed -- tools start working as their stage finishes. Returns: {"status": "complete" | "installing" | "failed", "message": "...", ...}

Convert GeoJSON to a Shapely geometry using shapely.geometry.shape. Args: geojson: GeoJSON dictionary. Returns: Dictionary with status, message, and geometry as WKT.

Convert a Shapely geometry (WKT) to GeoJSON using shapely.geometry.mapping. Args: geometry: WKT string of the geometry. Returns: Dictionary with status, message, and GeoJSON representation.

Get the area of a geometry. Args: geometry: WKT geometry. crs: Optional CRS of the geometry (e.g. 'EPSG:4326'). If geographic, the geometry is reprojected to UTM so the area is in square metres; if projected, the CRS's own unit is used; if omitted, the area is in the geometry's coordinate units (labelled 'unknown') -- WKT carries no CRS, so an unlabelled number could be square-degrees.

Return the most recent workflow provenance records from the audit log. Args: limit: Maximum number of records to return (newest last). Returns: Dict with status and a list of provenance records.

Get list of available CRS.

Get the bounds of a geometry.

Get the centroid of a geometry.

Check whether a local connector is currently connected to this server. A local connector must be running on the user's machine for ``request_local_file`` and ``list_local_directory`` to work. To start one, run on your local machine:: geoedge-mcp-local --server <SERVER_URL> \ --token <SHARED_TOKEN> \ --allow <LOCAL_DIR> Returns: connected (bool): True if a local connector is active. status (str): "connected" or "disconnected". message (str): Human-readable description.

Get the coordinates of a geometry (Point/Line/Polygon/Multi*/Collection).

Get information about a CRS.

Get geocentric CRS for given coordinates.

Get information about a geodetic calculation.

Get the type of a geometry.

Get full metadata for one registered layer by name. Includes 'missing', and 'open_in_editor'/'unsaved_changes' when the layer is currently open in the map editor.

Get the length/perimeter of a geometry. Args: geometry: WKT geometry. crs: Optional CRS of the geometry (e.g. 'EPSG:4326'). If geographic, the geometry is reprojected to UTM so the length is in metres; if projected, the CRS's own unit is used; if omitted, the length is in the geometry's coordinate units (labelled 'unknown') -- WKT carries no CRS, so an unlabelled number could be degrees.

Return a map-product template definition (branding + required elements).

Return the full definition of a named output schema (fields, types, CRS).

Return the project's GIS conventions: default working CRS, registered output schemas, available map templates, and the priority dataset names.

Retrieve the Coordinate Reference System (CRS) of a raster dataset. Opens the raster (local path or HTTPS URL), reads its DatasetReader.crs attribute as a PROJ.4-style dict, and also returns the WKT representation. A URL is checked before GDAL sees it: hosts resolving to private, loopback, link-local, reserved or multicast addresses are refused, using the same guard as ``fetch_dataset`` (opt out with ``GEOEDGE_MCP_FETCH_ALLOW_PRIVATE=1`` for an internal data server). That check is pre-flight only — GDAL follows redirects internally, so a hop from a public URL to a private one is *not* revali…

Retrieve taxonomic information for a given species name. Args: scientific_name: Scientific name of the species (e.g., "Puma concolor") Returns: Taxonomic info dict or error message.

Get UTM CRS for given coordinates.

Get UTM zone for given coordinates.

Compute Getis-Ord G for global hot spot analysis.

Local Getis-Ord G.

Run spreg.GM_Lag (spatial 2SLS / GMM-IV spatial lag model) on a cross-section. Returns coefficients with SE/z/p, fit metrics, (optional) AK test, and a small data preview.

Generate hillshade from a DEM raster. Args: raster_path: Path to the DEM raster. azimuth: Sun azimuth angle in degrees. angle_altitude: Sun altitude angle in degrees. output_path: Optional path to save the hillshade raster. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status, message, and output path if saved.

Load spatial data -- "load my data", "open this file", "read this shapefile", "add this spreadsheet of sites", "map these coordinates". Two ways in, and exactly one of them per call: - ``file_path`` -- a file on the **user's own computer**. This is the widest file reader on this server: more formats than ``read_file_gpd``, finds latitude/longitude or WKT columns in a plain CSV or an Excel spreadsheet on its own, and reads a zipped shapefile in place without unzipping it. Reach for this whenever the user hands over a file and it is not obvious what will read it. - ``content`` -- the…

Find intersection of two geometries. Args: geometry1: WKT geometry. geometry2: WKT geometry. crs1: Optional CRS of ``geometry1`` (e.g. 'EPSG:4326', 'EPSG:28355'). crs2: Optional CRS of ``geometry2``. When both are given and differ, ``geometry2`` is reprojected onto ``crs1`` first and the result is returned in ``crs1``. WKT carries no CRS, so when they are omitted the inputs are assumed to share one -- the response reports this via ``crs_status`` and warns if the coordinates say otherwise.

Check if a geometry is valid.

Global join counts for a BINARY (0/1) variable. Returns the BB (1-1), WW (0-0) and BW join counts, the permutation p-value for BB, and a chi-square test of independence. `dependent_var` must be binary; a continuous variable is refused with an explanation.

Local join counts for a BINARY (0/1) variable: one BB count per feature. `dependent_var` must be binary; a continuous variable is refused with an explanation. `p_values` is null for any feature that cannot have a BB join (a 0-valued feature, or one with no 1-valued neighbour) -- that is "not applicable to this feature", not a missing result.

Create a k-nearest neighbors spatial weights (W) object from point data. - data_path: path to point shapefile or GeoPackage - k: number of nearest neighbors - id_field: optional attribute name to use as observation IDs

List sub-layers exposed by an ArcGIS Enterprise Map/Feature Service. Args: connection: Named alias from list_db_connections() (recommended; must have driver: arcgis_rest). base_url: Raw Feature/Map Service root URL escape hatch, e.g. "https://gis.company.com/server/rest/services/Prod/MapServer". token: Raw token escape hatch, paired with base_url. Exactly one of connection/base_url must be given. Returns: {"status": "success", "layers": [{"id", "name", "type", "geometryType"}, ...]}

List configured named DB/service connection aliases (never returns secrets). Returns: {"status": "success", "connections": [{"name", "driver", "read_only", "description", "configured"}, ...]}

List the currently registered layers, most recent first. Each entry gives name, kind, CRS, extent and source tool, plus: 'missing' (true if the file it points at no longer exists), 'open_in_editor' and 'unsaved_changes' (present only for layers currently open in the map editor). A layer with unsaved_changes has edits that have not been written back to its source file yet -- offer to save before running analysis on it. The response is paginated. After months of use a registry can hold thousands of layers, and returning all of them in one tool result is large enough to crowd out the rest of th…

List files in a directory on the user's local machine. Useful for browsing what data files are available before fetching them with ``request_local_file``. The local connector must be running and the directory must be within its configured allowed roots. Args: path (str): Absolute path to a directory on the user's local machine. Example: ``"C:/GIS_Projects/survey_2024"`` Returns: status (str): "success" or "error". path (str): The directory path that was listed. count (int): Number of entries returned. entries (list[dict]): Ea…

List the available map-product (branding/layout) templates by name.

List all registered company output schemas by name.

List spatial (geometry-column) tables/views visible on a PostGIS connection. Args: connection: Named alias from list_db_connections() (recommended). connection_string: Raw postgresql:// DSN escape hatch for a one-off database that isn't pre-registered. Exactly one of connection/connection_string is required. schema: Postgres schema to inspect (default "public"). Returns: {"status": "success", "tables": [{"schema", "table", "geometry_column", "geometry_type", "srid"}, ...]}

List the registered priority datasets (name, source, category, default CRS).

List the available guided workflow templates. Returns: Dict with status and a list of {name, title, description} entries.

Make a geometry valid.

Merges two shapefiles with a database-style attribute join (not a spatial join). An explicit join key is required — either `on`, or both `left_on` and `right_on`. Args: shapefile1_path: Path to the left shapefile. Its geometry is preserved. shapefile2_path: Path to the right shapefile to merge (attributes only). output_path: Path to save the merged output shapefile. how: Type of merge. One of 'left', 'right', 'outer', 'inner'. Defaults to 'inner'. on: Column name to join on. Must be found in both shapefiles. left_on: Column name to join on in the left shapefile. rig…

Open a raster dataset in read-only mode and return metadata. This tool supports two modes based on the provided string: 1. A local filesystem path (e.g., "D:\Data\my_raster.tif"). 2. An HTTPS URL (e.g., "https://example.com/my_raster.tif"). The input must be a single string that is either a valid file path on the local machine or a valid HTTPS URL pointing to a raster. A URL is checked before GDAL sees it: hosts resolving to private, loopback, link-local, reserved or multicast addresses are refused, using the same guard as ``fetch_dataset`` (opt out with ``GEOEDGE_MCP_FETCH_ALLOW_PRIVATE=1`…

Get minimum rotated rectangle of a geometry.

Local Moran's I.

Compute Moran's I Global Autocorrelation Statistic.

Find the nearest point on geometry2 to geometry1 using shapely.ops.nearest_points. Args: geometry1: WKT string of the first geometry (e.g., a point). geometry2: WKT string of the second geometry. Returns: Dictionary with status, message, and the nearest point as WKT.

Normalize the orientation/order of a geometry using shapely.normalize. Args: geometry: WKT string of the geometry. Returns: Dictionary with status, message, and normalized geometry as WKT.

Min-max normalise a raster band to the 0-1 suitability range. Args: source: Path to the input raster. destination: Path to save the normalized single-band raster (float32). direction: "benefit" (higher value = more suitable) or "cost" (higher value = less suitable; the scale is inverted). band_index: 1-based band to normalize (default 1). overwrite: When False, refuse to run if destination already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dict with status, output path, and…

Safe MCP pipeline: Read shapefile, build/load W, convert numeric, check NaNs, run OLS. Returns the coefficients and R-squared, plus the spatial diagnostics the name promises: Moran's I on the residuals (statistic, z-score, p-value) and the LM/robust-LM error, lag and SARMA tests. Parameters: - data_path: path to shapefile or GeoPackage - y_field: dependent variable column name - x_fields: list of independent variable column names - weights_path: optional path to existing weights file (.gal or .gwt) - weights_method: 'queen', 'rook', 'distance_band', or 'knn' (used if weights_path not provide…

Open the user's native file manager at the folder containing a file this server already produced, with that file selected. Desktop app only -- there is no file manager to open on the hosted Railway server. Args: file_path: Path to a file already produced by another tool (e.g. the output_path from create_web_map or create_map). Absolute paths and paths relative to the configured storage folder are both accepted, but the resolved location must fall inside the storage root -- this tool launches a native OS process, so it confines the target even for c…

Overlay two GeoDataFrames using geopandas.overlay. Args: gdf1_path: Path to the first geospatial file. gdf2_path: Path to the second geospatial file. how: Overlay method ('intersection', 'union', 'identity', 'symmetric_difference', 'difference'). output_path: Optional path to save the result. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status, message, and output info.

Check if points are inside polygons using spatial join (predicate='within'). Args: points_path: Path to the point geospatial file. polygons_path: Path to the polygon geospatial file. output_path: Optional path to save the result. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status, message, and output info.

Project a geometry between CRS.

Select and rank the most suitable locations from a suitability raster. Picks either the highest-scoring ``top_n`` cells, or every cell whose value is >= ``threshold`` (when supplied), and writes them as ranked points to a GeoPackage. Cell centres are used as the point geometry. Returns the ranked results table for direct presentation to the user. Threshold mode can qualify a very large number of cells (a 10k x 10k surface has 100 million of them), so the number written is capped at ``max_sites`` and the number listed in the response at ``max_records``. Both caps are reported explicitly — the…

Perform algebraic operations (addition or subtraction) on two raster bands, handling alignment issues automatically. Both rasters are read onto raster1's exact grid, with each one's declared nodata excluded from the maths. A pixel that is missing in either input -- because it is fill, or because it falls outside the other raster's footprint -- is nodata in the result, not a half-sum: `raster1 + nothing` is unknown, not `raster1`. Parameters: - raster1: Path to the first raster (.tif). Its grid, CRS and extent define the output. - raster2: Path to the second raster (.ti…

Calculate min, max, mean, and std for each band of a raster. Parameters: - source: path to input raster (local or URL).

Evaluate an arithmetic/boolean expression across any number of rasters. This is the general map-algebra tool: use it for multi-criteria masks, indices, and scaled scores. Inputs are automatically aligned onto one reference grid before the maths runs, and each input's nodata becomes NaN so it propagates instead of contaminating the result with a sentinel value. Examples: * eligibility mask -> ``inputs={"rain": "rainfall", "slope": "slope_pct"}``, ``expression="(rain > 600) & (slope < 15)"``, ``dtype="uint8"`` * scaled index -> ``expression="M * clip(FPI / 25, 0, 1) * (1 - exp(…

Compute histogram of pixel values for each band. Parameters: - source: path to input raster. - bins: number of histogram bins.

Query one sub-layer of an ArcGIS Enterprise Feature/Map Service and materialize the result as a local vector layer. Uses the layer's /query endpoint with f=geojson, paginating via resultOffset/resultRecordCount until 'limit' rows are collected or the service reports no more results (exceededTransferLimit). Row count is always capped (default 5,000, hard max 50,000). Args: layer_id: Numeric sub-layer id (see list_arcgis_layers). connection: Named alias (recommended). base_url: Raw Map/Feature Service root URL escape hatch. token: Raw token escape hatch, paired with base_url. …

Reads a vector geospatial file and returns its CRS, bounds, columns, feature count and a 5-row preview. Args: file_path: Path to the vector file (.shp, .gpkg, .geojson, ...), or the name of a previously registered layer.

Read a PostGIS table or read-only SQL query into a materialized layer. Exactly one of 'table' or 'sql' must be given. Reads are always capped (default 5,000, hard max 50,000 rows) -- pass a smaller 'limit' for a quick preview. The result is materialized as a local GeoPackage via the same output_packaging pipeline every other read/write tool uses, and auto-registered as a named layer for downstream tools. 'sql' is written in PostgreSQL syntax with PostGIS spatial functions -- the target database's own dialect, NOT this server's expression dialect (the one ``filter_features`` and ``raster_expr…

Reclassify raster values using a mapping dictionary. Args: raster_path: Path to the input raster. reclass_map: Dictionary mapping old values to new values (e.g., {1: 10, 2: 20}). output_path: Path to save the reclassified raster. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status and message.

Register a vector or raster dataset under a short name so later tool calls can reference it by name instead of repeating the raw file path. CRS, feature/band count, and extent are probed automatically on a best-effort basis (registration still succeeds if the probe fails). Args: path: Path to the dataset (absolute, or relative to the storage directory) -- same rules as every other tool's path parameter. name: Optional name to register under. Auto-generated from the filename + timestamp if omitted. Must start with a letter or underscore; letters/digits/'_'/'-' …

Reproject a raster dataset to a new CRS and save the result. Parameters: - source: local path or HTTPS URL of the source raster. - target_crs: target CRS string (e.g., "EPSG:4326"). - destination: local filesystem path for the reprojected raster. - resampling: resampling method: "nearest", "bilinear", "cubic", etc. - overwrite: When False, refuse to run if destination already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior.

Fetch a file from the user's local machine and save it to server storage. The returned ``storage_path`` can be passed directly to ``read_file_gpd``, ``clip_raster``, and any other GIS tool that accepts a file path. **Shapefile note:** A ``.shp`` file requires companion sidecar files (``.dbf``, ``.shx``, ``.prj``, ``.cpg``). Call this tool once per sidecar file with the same base name, or convert the shapefile to GeoPackage (``.gpkg``) first for single-file transfer. Args: path (str): Absolute path to the file on the user's local machine. Example: ``"C:…

Resample a raster dataset by a scale factor and save the result. Parameters: - source: local path or HTTPS URL of the source raster. - scale_factor: multiplicative factor for width/height (e.g., 0.5 to halve resolution, 2.0 to double). - resampling: resampling method name: "nearest", "bilinear", "cubic", etc. - destination: local filesystem path for the resampled raster. - overwrite: When False, refuse to run if destination already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavio…

Rotate a geometry.

Run a guided workflow template, or preview it with dry_run. NOTE: deliberately registered with the plain @geoedge_mcp.tool() decorator, not @gis_tool() -- this function already builds and appends its OWN detailed provenance record (every step, not just a generic call summary). Wrapping it in @gis_tool() too would append a second, less detailed "direct_tool_call" record for the outer call right after the detailed one, breaking get_audit_log's "most recent record = the workflow that just ran, with a 'workflow' key" contract. NOTE (deploymentPlan.md Phase 2, F1): because this is the plain @geoe…

MCP Tool: Save any GIS-MCP result dict to files, only when the user requests. Spatial formats are written in the result's own CRS (``data["crs"]``, e.g. "EPSG:28356"), not silently converted to lat/long. The one exception is GeoJSON, which RFC 7946 defines as WGS84: its coordinates are reprojected to EPSG:4326 rather than merely relabelled. A result with no CRS is assumed to be EPSG:4326. Args: data: The dictionary returned by any GIS-MCP tool. filename: Base filename without extension. formats: List of formats to save (default = all, including a PDF report). folder: Target f…

Scale a geometry.

Simplify a geometry.

Spatial join between two GeoDataFrames using geopandas.sjoin. Args: left_path: Path to the left geospatial file. right_path: Path to the right geospatial file. how: Type of join ('left', 'right', 'inner'). predicate: Spatial predicate ('intersects', 'within', 'contains', etc.). output_path: Optional path to save the result. overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status, message, and output info.

Join each feature of the left layer to its NEAREST feature in the right layer (geopandas.sjoin_nearest), optionally within a maximum distance. Distance units follow the same locked policy as ``buffer_layer``: if the layers are geographic (lat/lon, e.g. EPSG:4326) both are reprojected to the appropriate UTM zone, the nearest-neighbour search and ``max_distance`` are evaluated in METRES there, and the result is returned in the original CRS. Searching in raw degrees would make "within 500 m" mean 500 *degrees* and can pick a genuinely more distant feature (a degree of longitude shrinks towards t…

Snap one geometry to another using shapely.ops.snap. Args: geometry1: WKT string of the geometry to be snapped. geometry2: WKT string of the reference geometry. tolerance: Distance tolerance for snapping. Returns: Dictionary with status, message, and snapped geometry as WKT.

Run giddy Spatial Markov on a panel (n regions x t periods) from a shapefile.

Find symmetric difference between geometries. Args: geometry1: WKT geometry. geometry2: WKT geometry. crs1: Optional CRS of ``geometry1`` (e.g. 'EPSG:4326', 'EPSG:28355'). crs2: Optional CRS of ``geometry2``. When both are given and differ, ``geometry2`` is reprojected onto ``crs1`` first and the result is returned in ``crs1``. Without them the inputs are assumed to share a CRS -- see ``crs_status``/``warning`` in the response.

Split a raster into square tiles of a given size and save them individually. Parameters: - source: input raster path. - tile_size: size of each tile (e.g., 256 or 512). - destination_dir: directory to store the tiles. - zip_output: when True (default), also bundle all tiles + a metadata.json into '<destination_dir>.zip' for a single-file download alongside the raw tile directory. Set False to skip zipping (or set GEOEDGE_MCP_DISABLE_AUTO_ZIP=1 to disable it server-wide; legacy GIS_M…

Transform coordinates between CRS.

Translate a geometry.

Create a triangulation of a geometry.

Create a union of multiple geometries. Args: geometries: WKT geometries. They must all be in the SAME CRS -- WKT carries none, so nothing in the data itself enforces this. Mixing, say, MGA (metres) and lat/lon (degrees) parts produces a valid-looking MultiPolygon with a nonsense extent and area. crs: Optional CRS shared by every input (e.g. 'EPSG:4326'). It is reported back as ``crs_used`` and is a declaration, not a conversion: reproject the inputs onto a single CRS before calling if they differ. Regardless of what is declared, the coor…

Combine two geometries. Args: geometry1: WKT geometry. geometry2: WKT geometry. crs1: Optional CRS of ``geometry1`` (e.g. 'EPSG:4326', 'EPSG:28355'). crs2: Optional CRS of ``geometry2``. When both are given and differ, ``geometry2`` is reprojected onto ``crs1`` first and the result is returned in ``crs1``. Without them the inputs are assumed to share a CRS: a union of mismatched inputs still looks like a valid MultiPolygon, so check ``crs_status``/``warning`` in the response.

Validate a tool-result dict against a named schema without writing files. Args: data: A GIS-MCP result dict (may carry a 'records'/'features' list, or be treated as a single feature record). schema_name: Registered schema to validate against. coerce: Apply type coercion / alias mapping during validation. Returns: Validation summary: valid flag, per-field/row errors, and warnings.

Show the user their spatial data on an interactive map in their own web browser. This is the tool for "show me", "visualise this", "map it", "plot these points", "display the result", "let me see it", "open this on a map". This is the ONLY tool for SHOWING data. ``create_web_map`` writes a standalone .html FILE for sending to somebody else — call that one only when the user asks to export, share, save or email a map. Accepts, in any combination: a registered layer name, an absolute or storage-relative file path, an http(s) URL to download, or a PostGIS table or query. **Local file paths are …

Create a Voronoi diagram from points.

Compute a weighted sum of all bands in a raster using specified weights. Parameters: - source: Path to the input multi-band raster file. - weights: List of weights (must match number of bands and sum to 1). - destination: Path to save the output single-band raster. - overwrite: When False, refuse to run if destination already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Nodata handling (T2-H7): a pixel that is nodata in *any* band has no weighted sum -- one of its terms is missing -- so it i…

Weighted-overlay multi-criteria suitability model over several raster layers. Each input layer is aligned to a common reference grid, min-max normalized to 0-1 (inverted for "cost" criteria), multiplied by its weight, and summed. Weights are automatically rescaled to sum to 1.0, so the output is a 0-1 suitability surface. Cells that are nodata in ANY layer are nodata in the output. Args: layers: List of raster paths, one per criterion (band 1 is used). weights: List of relative weights, one per layer (need not sum to 1). destination: Path to save the output suitability raster (fl…

Weighted multi-criteria scoring and ranking of vector features. Each criterion names an attribute field, a weight, and a type. The field is min-max normalized to 0-1 across all features ("cost" criteria are inverted), multiplied by its (auto-normalized) weight, and summed into a suitability score. Features are ranked best-first. Ideal for parcel / candidate-site selection over discrete units. Args: vector_path: Path to the input vector dataset (polygons or points). criteria: List of dicts, each: {"field": str, "weight": float, "type": "benefit"|"cost"}. "type" defaults to "be…

Create a spatial weights (W) from a shapefile using contiguity. - contiguity: 'queen' or 'rook' (default 'queen') - id_field: optional attribute name to use as observation IDs

Export a GeoDataFrame to a file (Shapefile, GeoJSON, GPKG, etc.). Args: gdf_path: Path to the input geospatial file. output_path: Path to save the exported file. driver: Optional OGR driver name (e.g., 'ESRI Shapefile', 'GeoJSON', 'GPKG'). overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status and message.

Write a registered layer or local vector file to a PostGIS table. Hard-refuses unless the resolved connection is explicitly write-enabled: a named alias must have read_only: false in db_connections.yaml, AND the caller must pass write=True -- a second, explicit opt-in beyond just calling this tool, mirroring plan/PLAN.md §8's "explicit opt-in for writes". The raw connection_string escape hatch is NOT accepted here -- only a named alias, since a raw connection string has no administrator-configured read_only flag to gate against (unlike reads, where the escape hatch has no destructive effect).…

Write a numpy array to a raster file using metadata from a reference raster. Args: array: 2D or 3D list (or numpy array) of raster values. reference_raster: Path to a raster whose metadata will be copied. output_path: Path to save the new raster. dtype: Optional data type (e.g., 'float32', 'uint8'). overwrite: When False, refuse to run if output_path already exists (instead of silently replacing it). Default True preserves the previous always-overwrite behavior. Returns: Dictionary with status and message.

Calculate statistics of raster values within polygons (zonal statistics). The vector layer is automatically reprojected to the raster CRS if they differ, so mismatched CRS inputs no longer produce silently wrong numbers. When either layer declares no CRS at all the alignment cannot be verified; that case is reported via `vector_crs_status`, not silently treated as a match. Args: raster_path: Path to the raster file. vector_path: Path to the vector file (polygons). stats: List of statistics to compute (e.g., ["mean", "min", "max", "std"]). Returns: Dictionary with status, mess…