MCP Server User Analytics #1236
Replies: 2 comments
-
|
One simple option for this might be passing the context around user intent or recent messages through your tool arguments. I think you could potentially provide instructions to the client/host in the tool description and/or in other instructions from the server. Note that since you'd be passing user data back to your system, you'd obviously need to ensure the user is clearly informed and that you're protecting any information being passed to a server this way. |
Beta Was this translation helpful? Give feedback.
-
|
Your question about understanding user journeys and deriving analytics from tool calls is being addressed in a broader proposal: Discussion #2369. It proposes an opt-in mechanism where the LLM client reports structured, anonymized feedback about its session-level experience — tool usability, reliability, documentation gaps, efficiency issues, and cross-tool interaction patterns. This would give server maintainers the kind of insight you were looking for without requiring ownership of the MCP client. There's a full SEP draft and a working prototype. Your perspective as a server maintainer would be valuable — would love your input. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Pre-submission Checklist
Question Category
Your Question
Hey guys, sorry if this is answered somewhere. I really tried to look (~30 mins)!
Maybe I'm a bit early to this issue, but I'm working on an MCP server and I'm trying to figure out how best to solve a user analytics problem. I'm trying to answer some questions about how to derive the user journey.
For example:
The only way to do this right now for an MCP Implementation would require "ownership" of the MCP Client and manually publish that data somehow to the MCP Server in a standardized way. Looking for any thoughts or suggestions!
Beta Was this translation helpful? Give feedback.
All reactions