Introduction

DeepSeek Harness (DSH) provides the ability for models to call external tools. However, for geospatial developers, directly connecting to local ArcGIS Pro often involves complex Python environment configuration and process management. The dsh-arcgis-pro-bridge plugin wraps ArcGIS-Pro-Bridge-MCP-Server, bridging ArcGIS Pro Python logic into DSH and allowing models to operate local GIS data directly through standard MCP tool names (such as mcp__arcgis__*).

What This Is

This is a DSH plugin maintained by y0urday. It uses DSH’s official @deepseek-ai/dsh-mcp-client to start a Python service via stdio, and leverages the Python subprocess included with ArcGIS Pro to execute ArcPy logic. The plugin is pure ESM JavaScript + vendored Python, with no build step, and supports direct installation from GitHub and installation via npm.

Core Features

  • Structure reading: Read the current project, layer list, and GDB structure.
  • Spatial analysis: Run basic tools such as Buffer and Clip.
  • Code execution: Support executing custom ArcPy code.
  • Tool exposure: Expose all capabilities to the model in the mcp__arcgis__* format.

Installation and Activation

It is recommended to install a specific version directly from GitHub to avoid build permission issues:

dsh plugin --profile web add github:y0urday/dsh-arcgis-pro-bridge#<commit-sha>

After installation, the arcgis-pro-bridge field should be visible in the configuration. Validate the configuration and fully restart dsh web:

dsh --profile web --dump-config

Typical Usage

After the plugin is loaded, the model can call the following tools:

  • Connectivity test: Call mcp__arcgis__ping to confirm the connection, then call mcp__arcgis__health_check and mcp__arcgis__doctor to confirm that ArcGIS Pro Python and ArcPy can be imported normally.
  • Data reading: Call mcp__arcgis__list_gis_layers to read the current project, or call mcp__arcgis__inspect_gdb to inspect the database structure.
  • Spatial processing: Call mcp__arcgis__buffer_features or mcp__arcgis__clip_features to perform analysis.
  • Custom code: Call mcp__arcgis__execute_arcpy_code to execute arbitrary ArcPy logic.

Applicable Scenarios and Notes

  • Environment requirements: Windows only, and ArcGIS Pro must be installed beforehand. A DSH version of 0.1.0-rc.6 or later is recommended, and Node.js >= 22.19. It is recommended to install uv or use Python 3.11+ with the mcp package already installed.
  • Operating mechanism: The service runs only on the local machine and does not open network ports. ArcPy is always executed in the Python environment included with ArcGIS Pro, so it does not pollute the Node environment.
  • Resource limitations: The official MCP bridge currently only bridges Tools. Upstream arcgis:// Resources are not registered. For read operations, use the same-name Tool (such as inspect_gdb).
  • Security notice: execute_arcpy_code is equivalent to executing code on the local machine. Enable it only on trusted machines and back up data before any write operations.

Brief Conclusion

By wrapping the MCP Server, this plugin lowers the barrier to using ArcGIS Pro in DSH. It is suitable for developers on Windows who need to embed GIS analysis capabilities into agent workflows. The full source code and directory structure can be found in the GitHub repository.