PyCaret: Open-Source Low-Code Machine Learning
pycaret.org · checked 15 Oct 2026
Low-code, open-source Python tooling for automating machine-learning experiments and deployment.
How it plugs in
Signal on Predictive Analytics Software7.7#1 of 37
PyCaret provides an open-source Python environment for low-code machine-learning workflows. Its experiment APIs handle classification, regression, clustering, anomaly detection, and time-series forecasting. You can compare and tune models, evaluate results, save portable pipelines, and serve predictions. Multi-user and deployment work can use a self-hosted React dashboard with a FastAPI backend.
Review
In a line
PyCaret combines a low-code Python library with a self-hosted platform for automated machine-learning workflows. It runs on Windows, macOS, Linux, the web, and self-hosted setups.
What it does
Classification, regression, clustering, anomaly detection, and time-series forecasting are available through its experiment APIs. These workflows fit experiments, compare models, tune settings, make predictions, and save or load models.
The default fitting pipeline splits data, fills missing numeric and categorical values, and encodes categories. Diagnostic charts cover model results, features, clusters, anomalies, forecasts, and exploratory analysis.
Classification experiments include measures such as accuracy, AUC, recall, precision, F1, kappa, and MCC. Other experiment types provide model families and adapters for their specific tasks.
How it connects
Installation uses PyPI and the command `pip install pycaret`; dashboard, explainability, and forecasting dependencies are optional extras. Its typed public API returns JSON data about models, metrics, and setup parameters.
Outputs can be stored as portable scikit-learn pipelines or, for time-series work, forecasting pipelines. FastAPI can serve a pipeline, while the optional PyCaret server exposes an HTTP prediction endpoint.
The optional self-hosted dashboard combines a React web interface with a FastAPI backend. Docker Compose can run the interface, backend, SQLite data volume, and local model files. Existing authentication can be connected, and the backend supports JWT-based access.
Good fit
Python users seeking a shorter route from data to tested machine-learning models may find PyCaret useful instead of assembling each workflow step themselves. It is useful for comparing registered candidates, tuning them, evaluating results, and deploying saved pipelines.
Teams working on forecasting, real-time deployment, or a self-hosted multi-user control layer may also fit this tool. The standard route uses a CPU; GPU estimators work when their required libraries are installed.
Watch for
In version 4.0, the former 3.x module-level functional API is gone. Existing 3.x code should stay pinned to the 3.x line, while new work should use the 4.0 object-oriented Experiment API; mixing the lines is unsupported.
The 4.0 repository labels that line as work in progress. Its changelog notes unfinished plot rewrites, legacy internals still being replaced, and no Python 3.14 support.
Check licensing for the part you deploy: the engine uses MIT, while control-plane components use BUSL-1.1 with a later Apache-2.0 change date and an additional-use grant.
Pricing
PyCaret is free and open source. No paid tier or enterprise plan is stated, and the offering has no free trial.
The PyCaret engine is MIT-licensed. The optional self-hosted control-plane components have the separate BUSL-1.1 licensing described in the project facts.
Verdict
PyCaret is worth considering when one Python toolkit should cover several machine-learning tasks, built-in comparison and tuning, portable pipelines, and optional self-hosted serving. Its API and dashboard options can support both individual experiments and shared deployment workflows.
Upgrading older projects requires care because the 4.0 API is not backward-compatible with 3.x. Teams needing Python 3.14 or finished 4.0 plotting work should check those requirements before adopting it.
Our take
PyCaret can suit Python teams seeking repeatable experiments across common machine-learning and forecasting tasks. Its free, open-source engine includes comparison, tuning, evaluation, pipeline saving, and deployment paths, while the optional server adds shared control features. Adoption needs more care for existing 3.x projects because the 4.0 API breaks compatibility. The work-in-progress status, unfinished plotting changes, Python-version limit, and separate control-plane license also matter when planning a production setup.
- Free, open-source engine with no paid tier stated
- Covers classification, regression, clustering, anomalies, and forecasting
- Compares and tunes models through experiment APIs
- Saves portable pipelines and supports FastAPI serving
- A self-hosted React dashboard and FastAPI backend are available
- PyCaret 4.0 breaks the older 3.x functional API
- The 4.0 line is labeled work in progress
- No Python 3.14 support is listed
- Some plot rewrites and legacy replacements remain incomplete
- Control-plane components use a separate BUSL-1.1 license
- Pick it ifPython users automating common machine-learning experiments; Teams needing self-hosted pipelines and prediction serving; Forecasting and model-comparison workflows
- Skip it ifYou must keep an existing 3.x functional-API project on the newest release; You require Python 3.14 support; You need a paid enterprise plan or finished 4.0 plotting features
Pricing
Free open-source software. The engine is MIT-licensed; no paid tier or enterprise plan is stated. Optional control-plane components have separate BUSL-1.1 licensing.
- PyCaretMIT-licensed open source; no paid tier or enterprise plan statedFree
- PyCaret open-sourceNo paid tier · MIT open-source engine · self-hostedFree
As pycaret.org lists them, checked 7 Oct 2026.
Questions
Is PyCaret free?
Yes. PyCaret is free and open source, with no paid tier or enterprise plan stated.
Which machine-learning tasks does PyCaret support?
Classification, regression, clustering, anomaly detection, and time-series forecasting are supported by its experiment APIs.
Can PyCaret deploy models?
Yes. Models can be saved as portable pipelines, served behind FastAPI, or deployed through the optional PyCaret server layer.
Does PyCaret have a web interface?
An optional self-hosted React dashboard with a FastAPI backend is available. Docker Compose can run the interface, backend, SQLite volume, and local model files.
What Python versions does PyCaret 4.0 support?
The installation information lists Python 3.11, 3.12, and 3.13. Python 3.14 is not supported.
Can older PyCaret 3.x code run on 4.0?
No. The 4.0 release removes the 3.x module-level functional API. Existing code should stay on 3.x, and mixing the two lines is unsupported.
Where can users get help?
The FAQ points users to GitHub Discussions for questions, GitHub Issues for bugs, and LinkedIn for announcements.