YavioYavio
Back to Resources

What is Yavio? The Analytics Layer for ChatGPT Apps and MCP Apps

Yavio is the first open-source analytics platform for ChatGPT Apps and MCP Apps. Capture tool calls, funnels, retention, and errors with a 3-line SDK integration. Self-hosted or Cloud.

You built an MCP server. You deployed a ChatGPT App. Users are calling your tools inside AI conversations — but you have no idea what's actually happening. Which tools get used? Where do users drop off? What errors are they hitting? You're flying blind.

Yavio fixes that. It's the first analytics and visibility layer built specifically for ChatGPT Apps and MCP Apps. Think of it as PostHog or Mixpanel, but purpose-built for AI-native applications.

The Problem: AI Apps Are a Black Box

When users interact with your MCP tools or ChatGPT App widgets, the AI platform sits between you and your user. You don't get pageviews. You don't get click events. Traditional analytics tools weren't designed for this.

What you're missing today:

  • Which of your MCP tools actually get called — and how often
  • Where users abandon a multi-step workflow inside a conversation
  • What errors your tools throw in production (and for which users)
  • Whether the changes you shipped last week improved anything
  • Which users are power users and which churned after one session

This isn't a nice-to-have. If you can't measure it, you can't improve it.

What Yavio Does

Yavio captures every interaction between AI platforms and your tools, then gives you a full analytics dashboard to make sense of it.

For MCP Apps, the SDK wraps your MCP server with a single function call. Every tool call, resource read, and prompt execution gets captured automatically — no manual event tracking required.

For ChatGPT Apps, the React hook tracks user interactions in your App widgets: button clicks, form submissions, navigation, conversions.

On the dashboard, you get an overview of all activity, per-tool breakdowns, funnels, user paths, retention curves, a live event feed, and error analysis. Everything you'd expect from a modern analytics platform, adapted for AI app workflows.

How It Works

Integration takes three lines of code. You install the SDK, wrap your MCP server with withYavio(), and you're done:

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

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

From that point, Yavio automatically captures every tool call, resource read, and prompt — with zero additional configuration. No manual event tagging, no code changes to your existing tools.

When you need more than automatic instrumentation, the SDK gives you explicit tracking methods: .identify() to tie events to known users, .step() and .track() for custom funnel stages, and .conversion() for business events like purchases or signups.

Open Source and Self-Hosted

Yavio is fully open source under the MIT license. The entire platform — SDK, ingestion API, dashboard, everything — lives in a single GitHub repository.

You have two deployment options. Yavio Cloud is the fastest path: install the SDK, run npx @yavio/cli init, and view your analytics at app.yavio.ai. The free tier includes 1 million events per month. Self-hosted gives you full control: clone the repo, run docker compose up -d, and the whole stack (ClickHouse, PostgreSQL, Next.js dashboard, Fastify ingestion API) runs on your infrastructure.

Both options use the same codebase. There's no feature gating, no "enterprise-only" dashboard. What you get in Cloud is exactly what you get self-hosted.

Key Features

Automatic MCP instrumentation. withYavio() proxies your MCP server and captures every interaction without you touching your tool handlers.

User identification. .identify(userId, traits) connects events to real users, enabling retention analysis, cohort breakdowns, and per-user debugging.

Funnels and conversions. Define multi-step funnels across tool calls and widget interactions. Track where users convert and where they drop off.

Error analysis. See which tools fail, how often, and for which users. Catch issues before your users report them.

PII stripping. Identity data is automatically removed before storage. Built-in privacy by design.

Workspace-based multi-tenancy. Teams, roles (admin, member, viewer), invitations, and multiple projects per workspace. Built for teams, not just solo developers.

Who Is Yavio For?

Yavio is for anyone shipping tools into AI platforms. If you've built an MCP server that connects to Claude, ChatGPT, Cursor, or any MCP-compatible client — Yavio shows you what happens after deployment. If you've built a ChatGPT App with interactive widgets — Yavio tracks how users engage with them.

This includes SaaS companies adding AI integrations, developer tool makers building MCP servers, agencies deploying AI workflows for clients, and indie developers who want to understand their users.

Get Started

Install the SDK and have analytics running in under five minutes:

npm install @yavio/sdk
npx @yavio/cli init

Or self-host the entire platform:

git clone https://github.com/yavio-ai/yavio-analytics.git
cd yavio-analytics && docker compose up -d

The dashboard is at localhost:3000. The documentation is on GitHub. The code is MIT-licensed.

You built the AI app. Now see what's happening inside it.