Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

Metasyn

github.com

  • Free plan
  • Windows
  • macOS
  • Linux
  • Self-hosted
7.4 Club score Visit github.com

What Metasyn is

Metasyn is a free, open-source Python package for generating synthetic tabular data from sensitive datasets. Users fit a MetaFrame to a dataframe, save the model and settings as auditable JSON, and then synthesize new rows. It handles multiple data types, missing values, categories, unique key columns, and structured strings such as postcodes or identifiers. Metasyn supports pandas and Polars dataframes and uses Faker to create localized names, emails, phone numbers, and other data. By default, its model excludes multivariate information to reduce disclosure risks. A disclosure-control plugin is available, but the project warns that synthetic output still needs human review and is not guaranteed to be free of sensitive information. Users can extend the package through plugins and export generated dataframes to CSV, Parquet, Excel, and other formats. It is self-hosted, with installation through pip or GitHub, and the project points to versioned Docker containers and a command-line interface. Metasyn is distributed under the MIT License.

Who it suits

Metasyn is intended for public organizations, research groups, and individual researchers who own sensitive datasets and need synthetic tabular data. It suits users comfortable with a self-hosted Python package and reviewing generated output.

Good to know, and what to watch

  • Handles missing values, categories, and structured strings.
  • Supports pandas and Polars dataframes.
  • Exports to CSV, Parquet, Excel, and other formats.
  • MIT License permits use, modification, and distribution.
  • Plugin system supports extensions.
  • Synthetic output requires human review.
  • Output is not guaranteed to exclude sensitive information.
  • Self-hosted package requires installation.

Metasyn

Free

Free plan

The verdict

Metasyn provides a configurable workflow for generating synthetic rows from tabular data, with dataframe integrations and multiple export formats. Its disclosure controls do not remove the need to review outputs for sensitive information.

What the maker publishes

Deployment
The project points users to versioned Docker containers and a command-line interface for building programs with Metasyn. source
Exports
Generated dataframes can be exported to CSV, Parquet, Excel, and other formats. source
Extensibility
A plugin system lets users or organizations add extensions to adapt Metasyn to their needs. source
Installation
The package can be installed with pip from PyPI or directly from GitHub for the development version. source
Integrations
Metasyn supports Polars and pandas dataframes and uses Faker to generate localized names, emails, phone numbers, and other data. source
Intended users
The documentation identifies public organizations, research groups, and individual researchers who own sensitive datasets as the intended audience. source
License
The software is distributed under the MIT License, which grants free use, modification, distribution, and sublicensing subject to its conditions. source
Privacy approach
By default, Metasyn excludes multivariate information from its model to reduce risks such as identity, attribute, or group disclosure. source
Privacy plugin
The maker offers a disclosure-control plugin, while warning that synthetic output still needs human review and is not guaranteed to be free of sensitive information. source
Purpose
Metasyn generates synthetic tabular data to help make sensitive datasets more accessible, reproducible, and reusable. source
Support
Users can report bugs or ask technical questions through GitHub issues and can contact Raoul Schram. source
Supported data
Metasyn handles multiple data types, missing values, categorical data, unique key columns, and structured strings such as postcodes or identifiers. source
Workflow
Users fit a MetaFrame to a dataframe, save its model and settings as auditable JSON, and synthesize new rows. source

Checked against the maker's pages on 4 Oct 2026.