Retentioneering MCP
MCP server and skills for product analytics: user paths, funnels, segments and transition graphs over event logs
- 946 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Runs analysis code and reads your event data locally
- May enable anonymous usage telemetry by default
Install
- Works in
Manual install
pip install retentioneeringInstalls the package with the engine and MCP server. Requires Python 3.10+.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs analysis code and reads your event data locally
- May enable anonymous usage telemetry by default
This is third-party code. Review the repository files before installing.
What it does
Retentioneering is a Python toolkit that loads an event log into an Eventstream object and builds transition graphs, step matrices, funnels and behavioral segments from it. A built-in MCP server exposes these tools to an agent: you ask in plain language and the agent calls specific tested analysis functions. The package ships agent skills with ready-made analysis recipes and rules, so the agent repeats reliable steps instead of inventing calculations. Data stays on your machine, the engine runs on DuckDB, and results export to a static HTML report.
Who it is for. For product analysts and managers who study user behavior from event data.
Good fit when
- You need to see where users get stuck or drop out of a funnel
- You need to compare paths of two cohorts or experiment groups
- You want the agent to compute transitions and segments over your log
Not a fit when
- You have no event data with user, event and timestamp columns
- You need only aggregate web stats without user-level sequences
Example request
Load my event log and show at which steps users most often drop off before purchaseMCP
- Transport
- sse
- Authentication
- not required
Limitations
The MCP server is beta and runs in a local Jupyter kernel; it does not work in cloud notebooks like Colab. Python 3.10+ is required. Anonymous usage telemetry may be on by default and can be disabled.
How to disable. Stop the MCP server and remove the package: pip uninstall retentioneering. Delete the copied skill folders from your agent's skills directory.
FAQ
Does data go to the vendor's server?
No, analysis runs in your environment and event data stays on your machine. Anonymous usage telemetry is separate and can be turned off.
How do skills differ from the MCP server?
Skills are analysis instructions and recipes for the agent, while the MCP server gives the agent direct access to analysis functions over your Eventstream.
Discussion
No questions or reviews yet.
Related
MCP serversJitsu MCP configures and debugs data pipelines: destinations, streams, functions, live events and syncs.
Dremio MCP server: an agent browses tables, looks for patterns in the data and analyzes cluster workload.
An agent queries GreptimeDB metrics, logs and traces with SQL, TQL and RANGE, with write protection and data masking
Official Tableau MCP server: the agent finds data sources, sends queries to them and opens dashboards and Pulse metrics
MCP server that connects Claude Code and Codex to an open R or Python session in RStudio and Positron, where your data lives
datawrapper-mcp: Datawrapper charts through an agent
datawrapper-mcp
An MCP server and app for Datawrapper: an agent creates, updates, publishes and exports charts to PNG from your data
README in short
The README presents Retentioneering as an open-source toolkit for reproducible product analytics on clickstream and event logs. Version 5 is rebuilt on a DuckDB engine with new widgets and adds an MCP server for agents. An analyst loads data into an Eventstream and gets interactive behavior graphs, step matrices, funnels and segments right in the notebook. Everything runs locally and reports export to HTML. Apache 2.0 licensed.