PM4Py

Overview

PM4Py brings process mining into Python code, making it a natural fit for developers, data scientists, and researchers who want to work directly with event data. Its breadth—from model discovery to object-centric analysis—is a strong draw, but the AGPL publication obligation can rule it out for proprietary deployments.

Key features

Discover and compare process models

PM4Py can derive directly-follows graphs, Petri nets, BPMN models, and process trees from event data. That choice of representations suits technical teams investigating processes from different angles. Replay and alignment-based measures compare observed behavior with process models, while root-cause analysis adds a way to investigate deviations. Data refresh is manual, so teams seeking automated recurring updates will need to manage that workflow themselves.

Work with event logs and multiple object types

The library imports and exports event logs, supports XES, and handles OCEL 1.0 and OCEL 2.0. Object-centric mining can analyze events associated with several object types—for example, orders, items, and deliveries—rather than forcing every analysis into a single-object view. That is a useful distinction for researchers and developers working with interconnected event data.

Connect mining to Python data science

PM4Py supports machine-learning feature extraction for use with pandas and scikit-learn, which makes it relevant when process analysis is part of a broader Python data-science workflow. Installation is recommended through pip, with Windows and Linux instructions. Visualization requires Graphviz to be installed and configured, an extra setup step that matters for users who need visual outputs.

Licensing and support

The community version is covered by AGPL v3: covered work using PM4Py must be published under AGPL when distributed or made available over a network. That is a consequential constraint for commercial teams building closed-source software. A commercial license permits closed-source use, and commercial evaluation without open-sourcing code is available by requesting a specific evaluation license. The community version lists no support; the commercial version includes dedicated professional support. Process Intelligence Solutions GmbH has been responsible for development, maintenance, distribution, and licensing since 2024.

Pricing

Community Version

free, billed free of charge. It is intended for teaching, research, and open-source projects under AGPL v3, with covered work published under AGPL. This is the straightforward choice for qualifying users who can meet the license terms; it is not a free route to closed-source commercial deployment.

Commercial Version

Custom pricing, by quote. It permits closed-source commercial use and includes regular releases, early feature access, and dedicated professional support. Commercial teams that need to keep their code proprietary should consider this option; those who only need to evaluate PM4Py commercially can request a specific evaluation license.

Platforms

PM4Py is available as an API for Linux, macOS, and Windows. It is self-hosted, so teams should expect to run it in their own environment rather than use a hosted application. Windows and Linux installation instructions are provided, and Graphviz configuration is required for visualization.

Who it's for

PM4Py suits developers, data scientists, and researchers who are comfortable building analyses in Python and want process discovery, conformance checking, or object-centric mining in their own code. It is less suitable for teams looking for a ready-made graphical application, automated data refresh, or proprietary use without a commercial license.

Pros and cons

• Pros: Multiple discovery outputs, including Petri nets, BPMN models, and process trees, give technical users options for representing event data.
• Pros: XES and both OCEL 1.0 and 2.0 support, alongside object-centric mining, suit work involving events tied to several object types.
• Pros: Feature extraction for pandas and scikit-learn connects process analysis with common Python machine-learning workflows.
• Cons: AGPL publication requirements can conflict with distributed or network-accessible closed-source products.
• Cons: Visualization depends on separately installing and configuring Graphviz.
• Cons: Manual data refresh means recurring analysis needs an update process managed by the team.
• Cons: The community version lists no support; professional support comes with the commercial version.

Alternatives

Compare process mining software if you want to survey options beyond a Python library. Mindzie Process Mining is worth considering if a desktop product with a free edition capped at 1,000 cases better suits your workflow. Celonis Process Mining Platform is another option, with a free plan and API, web, and Windows platforms.

ProcessMind may suit users seeking a web-based option with a free plan for one member, unlimited models, and 20 AI credits per month; its free plan does not include additional seats or viewer seats. ARIS is a paid alternative available through API, self-hosted, web, and Windows platforms. Cortado and RuM are free alternatives for Linux, macOS, and Windows.

Apromore Process Mining is worth considering if you need a paid option with a free plan and free trial, as well as API, self-hosted, and web platforms. Pega CRM is a paid web-based alternative with a free trial.

Verdict

Choose PM4Py if you want a capable process-mining toolkit inside Python and can work within its licensing terms. Its discovery, conformance, and object-centric capabilities make it especially useful for technical analysis of event data. Look elsewhere if you need a graphical, automatically refreshed workflow or must keep a distributed or network-accessible product closed-source without buying a commercial license.