retentioneering/retentioneering-tools

What it solves

Retentioneering provides a reproducible way to analyze clickstream and event log data. It replaces one-off scripts with tested analytical primitives, allowing analysts and AI agents to identify where users get stuck, discover behavioral segments, and understand the actual paths users take toward conversion or churn instead of relying on simple predefined funnels.

How it works

The toolkit uses a DuckDB-backed Eventstream object to process raw timestamped event data (from CSV, Parquet, or databases). It offers a chainable API for data processing—such as sessionization and event filtering—and renders interactive widgets (Transition Graphs, Step Matrices, Sankeys) directly in Python environments like Jupyter or VS Code. It also includes an MCP server that allows LLM agents to run these analyses and generate validated reports.

Who it’s for

It is designed for product analysts, researchers, and developers who want to perform deep behavioral analysis on user event data without needing a hosted SaaS platform.

Highlights

• Multi-resolution analysis: Ability to zoom from individual events to long-term customer journeys.
• Interactive Widgets: Visual tools for transition graphs, cluster analysis, and funnel tracking.
• Diff Mode: Overlays two different user segments to visually compare behavioral differences.
• MCP Server: Enables AI agents to explore eventstreams and build auditable reports.
• Privacy-focused: Runs locally in the user's own environment; raw data does not leave the machine.