ext-apps/examples/cohort-heatmap-server
Repository files (latest commit first)
Filename Latest commit message Latest commit date
Olivier Chafik 3ac7c7935b
Add RESOURCE_MIME_TYPE constant (#108)
* Add RESOURCE_MIME_TYPE constant

* Update app-bridge.ts

* update syntax
2025-12-09 19:10:56 +01:00
..
src Use content instead of structuredContent in examples 2025-12-05 16:35:20 -06:00
mcp-app.html Add cohort-heatmap-server example 2025-12-04 14:32:53 -06:00
package.json Smoothen the process of getting started (#103) 2025-12-09 16:15:32 +01:00
README.md Smoothen the process of getting started (#103) 2025-12-09 16:15:32 +01:00
server.ts Add RESOURCE_MIME_TYPE constant (#108) 2025-12-09 19:10:56 +01:00
tsconfig.json Add cohort-heatmap-server example 2025-12-04 14:32:53 -06:00
vite.config.ts Add cohort-heatmap-server example 2025-12-04 14:32:53 -06:00

Example: Cohort Heatmap App

A demo MCP App that displays cohort retention data as an interactive heatmap, showing customer retention over time by signup month.

Initial view Hover state Low retention hover

Features

  • Cohort Retention Heatmap: Color-coded grid showing retention percentages across cohorts and time periods
  • Multiple Metrics: Switch between Retention %, Revenue Retention, and Active Users
  • Period Types: View data by monthly or weekly intervals
  • Interactive Exploration: Hover cells for detailed tooltips, click to highlight rows/columns
  • Color Scale: Green (high retention) through yellow to red (low retention)
  • Theme Support: Adapts to light/dark mode preferences

Running

  1. Install dependencies:

    npm install
    
  2. Build and start the server:

    npm run start:http  # for Streamable HTTP transport
    # OR
    npm run start:stdio  # for stdio transport
    
  3. View using the basic-host example or another MCP Apps-compatible host.

Architecture

Server (server.ts)

Exposes a single get-cohort-data tool that returns:

  • Cohort rows with signup month, original user count, and retention cells
  • Period headers and labels
  • Configurable parameters: metric type, period type, cohort count, max periods

The tool generates synthetic cohort data using an exponential decay model with configurable retention curves per metric type.

App (src/mcp-app.tsx)

  • Uses React for the heatmap visualization
  • Fetches data on mount and when filters change
  • Displays retention percentages in a grid with HSL color interpolation
  • Shows detailed tooltips on hover with user counts and exact retention values
  • Supports row/column highlighting on cell click