Community project: driving ArcGIS Pro from an AI agent (ArcPy CLI + live MCP bridge) — is agent/MCP integration on the team's radar?

[Jasper0122]

Hi ArcGIS API for Python / GeoAI team 👋

I built an open-source bridge that lets an LLM agent (e.g. Claude) operate ArcGIS Pro end to end, and wanted to share it with the team that's closest to Esri's GeoAI work — and ask whether agent/MCP integration is something you're exploring.

Repo: https://github.com/Jasper0122/CLI-Anything-Arcgis-Pro

It has two modes:

• Headless CLI over ArcPy — info / project / layout (incl. Map Series) / data / gp (any geoprocessing tool) / batch, JSON in / JSON out, with self-dispatch into Pro's arcgispro-py3.
• Live bridge + MCP — an in-process ArcGIS Pro SDK (.NET 8) add-in exposes the open Pro session over a local socket, wrapped as MCP tools (ping / zoom_to / query / run_gp / export_layout), so an agent drives the running project and you watch it happen in the UI.

I know this repo is the Web GIS arcgis package rather than ArcPy / the Pro SDK — I'm posting here because the audience and the GeoAI direction overlap, and I'd value the team's perspective.

A few open questions:

1. Is agent / MCP / tool-calling integration with ArcGIS something the team is looking at?
2. If so, is there a preferred direction the community should build toward (so efforts align rather than fragment)?
3. Any interest in feedback or a community contribution on this front?

(Not affiliated with Esri; ArcGIS, ArcGIS Pro and ArcPy are trademarks of Esri. This is an independent community integration.)

Happy to demo or move this to Esri Community if that's a better venue. Thanks!

[nanaeaubry]

Thank you for sharing this here, we will transmit it to the teams associated and comment here any updates