Execute Python code in the persistent ArcPy session.

arcpy is already imported. Variables persist between calls, so you can build up state across multiple invocations. If the last statement is a bare expression its repr is returned (like a REPL). stdout is captured.

Use this for anything ArcPy can do: geoprocessing, arcpy.da cursors, arcpy.mp map automation, raster algebra, arcpy.Describe, etc.

Run a geoprocessing tool by name and return its outputs and messages.

tool accepts either the flat name ('Buffer_analysis') or module form ('analysis.Buffer'). Positional parameters go in args, named parameters in kwargs. Use describe_gp_tool first if unsure of the signature.

Search all available geoprocessing tools (including extension toolboxes).

Space-separated terms are ANDed, matched against tool names like 'Buffer_analysis'. Empty query lists everything (up to limit).

Get the syntax and full documentation for a geoprocessing tool.

Create vector data (shapefile or geodatabase feature class) from GeoJSON.

This is the fastest way to MAKE data from scratch: emit a GeoJSON FeatureCollection and pick an output path. output_path decides the format: C:\data\roads.shp creates a shapefile, C:\data\my.gdb\roads a file-geodatabase feature class. Attribute fields are created from the feature properties automatically. Coordinates must be WGS84 lon/lat (the GeoJSON spec); use run_gp_tool 'Project_management' afterwards if you need another CRS. Existing outputs are overwritten. The geometry type is inferred from the data; set geometry_type (POINT, MULTIPOINT, POLYLINE, POLYGON) only when the collection mixes types and you must pick one. Returns a description of the created dataset.

Read vector data (shapefile, feature class, layer) as GeoJSON.

The inverse of create_features — lets you inspect actual geometries and attributes. Optional where is a SQL where-clause to filter rows (e.g. "POP > 10000"); limit caps returned features (response says if truncated). Output coordinates are WGS84.

Describe a dataset: type, spatial reference, extent, fields, row count.

Works on feature classes, shapefiles, rasters, tables, workspaces — anything arcpy.Describe understands.

Inventory a workspace (geodatabase or folder): feature classes, rasters, tables and datasets it contains.

Inspect an ArcGIS Pro project (.aprx): maps, layers, data sources, layouts and default geodatabase.

Execute Python INSIDE the currently open ArcGIS Pro application.

Unlike arcpy_execute (headless background session), this runs in the live Pro session the user is looking at: arcpy.mp.ArcGISProject("CURRENT") works, added layers appear immediately, the open project can be saved. Requires the user to have cowork mode running — if no listener responds, the error includes the exact one-liner they must paste into Pro's Python window. Variables persist between calls (separate namespace from arcpy_execute). Caution: prefer data/layer/symbology operations; avoid rapid or repeated view/camera manipulation, which can destabilize Pro. Pass action as a short human label ("Applying symbology") - it shows in the user's add-in activity log. House rules: don't create new .aprx files, maps, layouts or layers unless the user asked; after changing a map or layout, call its .openView() so the change appears instantly.

Check the ArcPy session: license level, workspace, live variables. Starts the session if it isn't running yet (first start is slow).

Restart the ArcPy worker process. Clears all session variables and releases any locks/licenses. Use after a hang, crash, or to free state.