ArcGIS Pro MCP
Provides an ArcGIS Pro add-in that exposes native ArcGIS Pro tools, geoprocessing tools, and optional arcpy/Python execution over MCP, allowing AI clients to work directly with the currently open ArcGIS Pro project, including its data, maps, and layouts.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
ArcGIS Pro MCP · 优化版(AgentKit)
An optimized, open-source ArcGIS Pro MCP distribution — one repository, three parts: the MCP add-in, an on-demand dialog guard, and on-demand template distillation.
Upstream project & original author
The MCP module is built on Knight60/ArcGIS-Pro-MCP, originally created by Pisut Nakmuenwai, and licensed under AGPL-3.0-or-later. This project carries local patches up to v1.1.4 and adds the guard/distill modules. Upstream attribution is preserved in NOTICE and in the source snapshot shipped with the customer package.
What it is
Let an AI assistant (Claude, Codex, or any MCP client) work on the ArcGIS Pro project already open in front of you — your data, your maps, your layouts. No copies, no exports, no cloud round-trip.
Part | What it does | Status | |
A | MCP | ArcGIS Pro add-in: 113 native tools (in-process, ~9 ms), every geoprocessing tool, optional arcpy via RunCode / Python bridge. Listens on | Customer-ready v1.1.4 on the |
B | Guard | On-demand dialog guard for ArcGIS Pro's "third-party code library" trust dialog. Nothing runs in the background; the guard is summoned only when an operation may raise the dialog, then exits. | Skeleton — design in docs/design-2026-11-16.md §3 |
C | Distill | On-demand template distillation. Turn a sample map figure into a reusable layout spec: deterministic measurement, assistant-guided semantic pass, validation, preview, review — one job, no daemon. | Skeleton — design in docs/design-2026-11-16.md §4 |
Why "on demand"
No resident processes. When no risky operation is running, there is no guard process and no watcher polling in the background.
The guard lives for one operation. It is summoned right before an operation that may trigger the trust dialog, and exits after the operation settles. Auto-click success is invisible; only a failure raises the topmost alert window (≥30 s, audible). It only ever clicks the third-party-code trust dialog — never overwrite/delete confirmations.
Distillation is a one-shot job. No inbox watcher, no service. You run it when you need it; every step is idempotent and resumable.
flowchart LR
AI[AI client<br/>Claude · Codex · others] -->|MCP over HTTP<br/>127.0.0.1:6520| M[ArcGIS Pro MCP add-in<br/>inside ArcGIS Pro]
M -.->|summon only when needed| G[Dialog guard<br/>short-lived process]
M -->|optional arcpy| R[RunCode toolbox /<br/>in-Pro Python bridge]
F[Sample map figure] -->|one-shot job| D[Template distillation]
D -->|layout spec| MRelated MCP server: ArcGIS Pro MCP
Quick start — customer edition (v1.1.4)
The ready-to-use package — add-in, stdio bridge, RunCode, dialog guard, Python bridge, installers, docs, and the full AGPL source snapshot — is published on its own branch:
Browse:
customer-v1.1.4Download: ZIP archive
Unzip anywhere and run
install/00-install.cmd(no administrator rights needed; requires ArcGIS Pro 3.7+).Open ArcGIS Pro; when asked to trust the unsigned add-in, choose Yes. Confirm the MCP tab shows running (port
127.0.0.1:6520).Connect your AI client (see below).
Run
install/01-check.cmd— all items[OK]means you are ready.
Nothing needs Python, pip, or an internet connection: the add-in runs on the .NET runtime that ships with ArcGIS Pro 3.7+, and the stdio bridge and guard use the PowerShell that ships with Windows. The package's SHA256SUMS.txt covers every file.
Connect an AI client
Client | How |
Claude Code |
|
Claude Desktop | Run |
Codex | Add |
Other MCP clients | Streamable HTTP to |
Repository layout
├── main (this branch) project skeleton: guard + distill modules, design docs
├── customer-v1.1.4 ready-to-use package (add-in 1.1.4 + everything above)
├── src/arcgis_agentkit/ guard / distill Python packages (CLI skeletons)
├── skills/ assistant-facing skills (template-distill, arcgis-pro-mapping)
├── library/ template library schema (v1) and examples
├── docs/ design & architecture notes
└── tests/ pytest smoke testsDevelopment install:
python -m pip install -e .
arcgis-guard doctor
arcgis-distill doctorRoadmap
M1 — Guard v1: summon protocol, event-first UIA detection, add-in auto-arm around
run_python_toolbox_tool, Pester tests.M2 — Distill v1: import / measure / validate / preview / library,
template-distillskill, template MCP tools.M3 — Open-source release: versioning, bilingual docs, CI, release assets, clean-machine walkthrough.
Details, acceptance criteria and trade-offs: docs/design-2026-11-16.md.
License, credits & disclaimer
License: AGPL-3.0-or-later. Because the MCP module is derived from an AGPL-3.0-or-later project, the combined repository is AGPL-3.0-or-later as well.
Upstream project & original author: Knight60/ArcGIS-Pro-MCP by Pisut Nakmuenwai — thank you. Local modifications are declared in NOTICE, RELEASE-NOTES.md, and the source snapshot (
source/SOURCE-NOTICE.mdin the customer package).Trademark: not affiliated with Esri. "ArcGIS" is a trademark of Esri. This is a third-party tool.
Privacy / network: all services bind to
127.0.0.1only; no telemetry, no cloud calls.Maintainer: 柳皓文 (@XNYGggg).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Connect AI assistants to your approved computer for files, commands, browser and desktop tasks.
GIS tools for AI agents: 65 free tools + 8 paid (hazard/site-scouting/GeoJSON export)
Connect your AI assistant to Blender for 3D modeling, Geometry Nodes, materials, lighting, animation, rendering and export. Inspect real scenes and viewport images, refine the result and restore checkpoints. OAuth connects your chat to Blender Bridge Desktop on an online Windows 10/11 x64 computer running Blender 4.2+. The chat can run on another device. Requires an active Blender Bridge account; 7-day trial available. AI model access is separate.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Related MCP Servers
- AlicenseAqualityCmaintenanceEnables AI assistants to control ArcGIS Pro through natural language, executing geoprocessing tools, editing data, and automating GIS workflows via a persistent ArcPy session.121Apache 2.0
- AlicenseBqualityBmaintenanceLets AI assistants control a live ArcGIS Pro session through arcpy, including layer management, attribute queries, geoprocessing, symbology, and export.11269AGPL 3.0
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to perform spatial analysis in ArcGIS Pro using natural language by exposing all ArcGIS Pro toolboxes as MCP tools.11MIT
- FlicenseAqualityDmaintenanceEnables natural-language control of a live ArcGIS Pro session, including layer management, geoprocessing, feature querying and editing, map layout export, and publishing hosted feature services.31-