MLJAR AutoML

MLJAR AutoML gives Python users a configurable route through supervised machine learning on tabular data. It is most useful for classification and regression workflows where model evaluation and reusable experiment summaries matter. Its main boundary is firm: it is not built for forecasting or time-series modeling.

Overview

You can run MLJAR AutoML in MLJAR Studio notebooks or use its open-source Python package in scripts and notebooks. Depending on the selected mode and settings, it can preprocess data, select and tune models, engineer and select features, and create ensembles. That can automate a substantial part of a modeling workflow, but users remain responsible for choosing algorithms, validation, metrics, and a time budget appropriate to the task.

Experiments can provide model comparisons, validation results, training times, explanations, and reusable saved artifacts. Structured summaries can be produced in Markdown and JSON. The reports and explanations vary with the task, mode, algorithms, and explanation settings, so the output is not identical for every run.

Key features

Four configurable training modes

Explain, Perform, Compete, and Optuna offer different modes of training, with configurable algorithms, validation, metrics, time budget, and explanations. This makes the tool adaptable to different modeling priorities; it also means users need enough machine-learning judgment to set up a meaningful comparison rather than relying on one fixed recipe.

Fairness-aware evaluation

Fairness-aware training evaluates models against a chosen metric and threshold and can try sample reweighting to mitigate disparities. That provides a way to bring fairness considerations into a workflow, but it does not guarantee a fair model; users should treat it as an evaluation and mitigation aid, not a certification.

Prediction apps and deployment

The Python package integrates with Mercury to generate apps for single predictions and batch CSV predictions. Those apps can be previewed locally, published through MLJAR Platform, or self-hosted with Docker. The deployment path is useful when a trained model needs a simple prediction interface, though it centers on generated Mercury apps rather than a general forecasting workflow.

Pricing

The mljar-supervised Python package is free and open source under the MIT license. It is the direct choice for users who want the AutoML package itself without a subscription. The separately priced Mercury plans add prompt, published-conversation, and public or private web-app allowances; those quotas apply to Mercury, not to the package's training modes.

• Free: free, with 50 prompts / month, 10 published conversations, and 1 public Mercury web app. It gives a limited way to use Mercury, with no private app allowance stated.
• Pro: $20 a month, with 500 prompts / month, 50 published conversations, 3 public Mercury web apps, and 1 private Mercury web app. This suits an individual or small project that needs more Mercury capacity and a private app.
• Business: $60 a month, with 2,000 prompts / month, 200 published conversations, 10 public Mercury web apps, and 3 private Mercury web apps. It fits heavier Mercury use or a larger set of published apps.
• Perpetual license: $199, paid once, billed as one payment, yours forever, with 1 year of updates, local LLMs with Ollama, and bring your own provider keys. This is a separate Mercury license option, not a replacement price for the open-source package.

The free package is the lowest-cost way to run AutoML, while Mercury's free tier is capped at 50 prompts / month and 1 public app. There is no free trial. The published plan terms give quotas and update coverage where stated, but no seat limits or renewal terms to weigh.

Platforms

MLJAR AutoML supports Linux, macOS, and Windows, and can be used through web and self-hosted options. The Python package, notebook workflow, and Docker self-hosting make it relevant to users who want to work locally or manage deployment themselves. MLJAR's privacy policy describes SSL/TLS encrypted connections, access controls, and authentication mechanisms; it also advises against submitting sensitive or regulated data to AI features unless authorized.

Who it's for

This is a good fit for analysts and Python users working on supervised tabular classification or regression who want automated preprocessing, model selection, tuning, and experiment comparison without surrendering control over training settings. It is less suitable for teams whose core need is time-series modeling, forecasting, or a fairness guarantee. MLJAR identifies Aleksandra Płońska as Founder and Executive Director and Piotr Płoński as Founder and CTO; its About page invites users to contact the team with questions.

Pros and cons

• Pro: Four configurable training modes let users align workflows with different modeling goals instead of accepting a single fixed pipeline.
• Pro: Model comparisons, validation results, timings, explanations, and saved artifacts make experiments easier to review and reuse.
• Pro: The open-source Python package costs nothing and supports both scripts and notebooks.
• Con: Its scope is supervised tabular data; forecasting and time-series modeling are not supported.
• Con: Report and explanation coverage depends on task and configuration, while users must still choose appropriate validation and metrics.
• Con: Fairness-aware evaluation can attempt mitigation but cannot ensure that the resulting model is fair.

Alternatives

For a free open-source Python library, AutoGluon is a relevant alternative. If you want a free, open-source visual analytics platform, consider KNIME Analytics Platform; Orange Data Mining is another free software option. For forecasting specifically, Prophet is an open-source forecasting tool with R and Python implementations, a closer fit than MLJAR AutoML's tabular classification and regression scope.

Weka, Aible, and Amazon SageMaker Autopilot are also alternatives to consider. ZebraDesigner is another listed option.

Browse more tools in AutoML Software or Predictive Analytics Software.

Verdict

Choose MLJAR AutoML if you need configurable supervised learning for tabular classification or regression, with experiment reports and a free open-source Python package. Its combination of automation and adjustable training settings is the reason to choose it. Look elsewhere if your work depends on forecasting or time-series models, or if you need fairness guarantees rather than an evaluation and mitigation attempt.