How much traffic?
Count MCP requests over time and by method. Notifications are counted separately, so they do not inflate request totals.
MCP Analytics for GA4
Track your MCP server in Google Analytics.
See which tools get called, which AI assistants use them, and how often. Add an open source Java library to your server. The reports live in your own GA4 account.
Open sourceJava 17+Apache 2.0
Your MCP serverJava libraryYour GA4 account
Know how your server is used.
Keep MCP activity alongside the rest of your analytics, without running another dashboard or hosted collector.
Which tools?
Break down tool calls by name. Track outcomes and duration to understand what gets used and where calls fail.
Which AI assistants?
Compare reported agents such as Claude, ChatGPT and Cursor, with recognized bots and health checks identified separately.
Where did they come from?
Record explicit source and campaign tags or available referrer evidence. When discovery information is missing, it stays unknown.
Agent names are reported metadata, not verified identities or visitor counts. Optional application hooks also record searches, link opens and attributed orders. See metric coverage.
A dependency.
A small request hook.
Use your existing Java MCP server. Events leave through a bounded background queue, keeping network calls off the request thread.
Follow the Google Analytics setup guideInstall with Gradle
repositories { mavenCentral() }
dependencies {
implementation 'io.github.shanifdhanani:mcp-analytics-ga4:0.1.1'
}- Configure your GA4 measurement ID and private API secret on the server.
- Add the request hook, then verify received events in Google Realtime.
- Create the saved reports using the guide, or give it to Claude or ChatGPT with authorized Chrome access.