Post

Mike Sullivan
@analyticsedge.com

Experimental...a local GA4 MCP server with a simple installer that doesn't require the AI to figure out how to work with the APIs. Premise: AI's shouldn't be forced to work with raw API methods. Using AE connector tech should dramatically reduce the effort to query and process the responses.

0 reposts0 likes

Mike Sullivan
@analyticsedge.com

Thoughts so far: - it's not significantly faster, although no direct comparison tests have been run yet -- I want to compare tokens used; they should be a lot lower. - getting an AI to NOT use the API terminology was surprisingly difficult --I had previously installed the official analytics-mcp.

Mike Sullivan
@analyticsedge.com

- installation is a breeze. No command line mumbo-jumbo. Simple logins avoid the whole project/credential file nightmare. - building a local MCP server with a user interface opens up a lot of possibilities.

Mike Sullivan
@analyticsedge.com

- does the data have to be sent back to the AI, or could it be displayed/saved locally for privacy and use the AI just for query generation? - more connector functionality to be exposed, but need to think about how to implement it to make life easier for the AI. ....so many ideas...