Open-source product analytics for ChatGPT Apps and MCP Apps

ChatGPT Apps SDK
1

Hey everyone,

We’ve been building ChatGPT Apps with the Apps SDK and ran into a problem that I think many of you are dealing with too: once your app is live, you have almost no insight into how people actually use it.

We wanted the same kind of product analytics that web apps get from PostHog, but for ChatGPT Apps. So we built Yavio, an open-source (MIT) product analytics SDK that wraps your MCP server and captures everything automatically.

How it works

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { withYavio } from "@yavio/sdk";

const server = withYavio(new McpServer({
  name: "your-chatgpt-app",
  version: "1.0.0"
}));

That’s it. From here, Yavio automatically captures:

  • Tool call analytics: which tools get called, how often, by whom, with what parameters
  • Latency per tool: which tools are fast and which are slow
  • Error tracking per tool: not just “did my server crash” but which specific tools fail, how often, and with what inputs
  • Session funnels: which sequences of tool calls lead to successful outcomes vs. drop-offs
  • Retention: are users coming back to your app after day 1, day 7, day 30?
  • Revenue attribution: if you’re monetizing, connect usage patterns to revenue

For ChatGPT Apps with interactive UI, we also have a React hook (useYavio()) that captures widget interactions alongside tool calls, so you get the full picture.

Links

Would love feedback from anyone building ChatGPT Apps. What product analytics are you missing? What questions do you wish you could answer about your app’s usage?