Open Wrangler is a visual dataframe editor for VS Code and editors based on it. Open files or live notebook data, make changes visually, and keep the generated Python or R code.
Explore rows, profiles, filters, and cleaning history in one workbench.
- Stable: choose Install on the Visual Studio Marketplace or Open VSX, or download the latest GitHub release.
- Preview: choose Install Pre-Release Version on the editor listing, or download a preview from GitHub prereleases.
For a downloaded VSIX, open the Extensions view and choose Views and More Actions → Install from VSIX….
1. Open a source. For CSV, TSV, Parquet, JSONL/NDJSON, or Excel, choose Open in Open Wrangler from Explorer, an editor tab, or the editor toolbar. In a Python notebook, run a supported dataframe and choose Open in Open Wrangler on its output or from the notebook toolbar.
Open a live notebook dataframe from its output and continue in the workbench.
2. Preview a transformation. Choose Add step, select an operation, and configure it. Changed values and generated code appear as a draft. Filters and sorts affect only the current view; they do not become cleaning steps.
Preview changed values and generated code before applying a step.
3. Apply the step and keep the code. Choose Apply step, then copy the generated code or use Open Wrangler: Export Generated Script. Supported notebook and R-document sessions can insert code into the document that opened the dataframe. Cleaned-data export always asks for a separate CSV or Parquet destination; Open Wrangler never overwrites the source.
Opening data, running code, or exporting requires a trusted workspace. Open Wrangler stays inactive in Restricted Mode.
- Explore a paged grid with column profiles, dataset summaries, typed filters, ordered sorts, full-schema search, rectangular selection, and copy.
- Clean rows, columns, types, text, categories, numbers, dates, missing values, and reshaped data. Drafts can be applied or discarded, and applied steps can be inspected, edited, deleted, or undone.
- Keep executable code for the selected engine beside the preview. Viewing filters and sorts remain separate from the cleaning plan and exported result.
- Open delimited text, Parquet, JSON Lines, and Excel files, or continue from live notebook and interactive dataframes.
| User action | File sessions | Notebook and interactive sessions |
|---|---|---|
| Open and view | CSV, TSV, Parquet, JSONL/NDJSON, and Excel; Open Wrangler selects a compatible backend | Pandas, Polars, DuckDB relations, local PySpark batch dataframes, and R frames |
| Filter, sort, profile, and copy | Available | Available |
| Preview and apply cleaning steps | Pandas and Polars; experimental DuckDB file editing | Pandas and Polars; selected R workflows. DuckDB relations and PySpark remain view-only |
| Copy, save, or insert code | Copy or save generated Python code | Copy or save generated Python or R code; insert it only into the notebook or managed document that opened it |
| Export cleaned data | Write CSV or Parquet to a separate destination | Available from supported editing sessions; view-only sessions cannot export |
The generated reference lists every command, setting, operation, and supported parameter.
Open Wrangler requires VS Code 1.106 or newer. VS Code is the primary target. On Linux, Cursor is one tested compatibility example for installation, activation, and representative grid, cleaning, export, and recovery flows. It is not tested across every VS Code feature. Support for other VS Code-based desktop editors is experimental. Browser-hosted editors are unsupported, and Remote SSH is outside the current compatibility coverage. Python file and notebook workflows use Python 3.10 through 3.14. If a required package is missing, Open Wrangler names it and asks before installing anything.
R workflows open data.frame, tibble, and data.table values. IRkernel works in VS Code on Linux, macOS,
and Windows, and in Cursor on Linux. Selected R terminal workflows are available on Linux. Direct .R, .Rmd, and
.qmd execution is available on macOS and Linux, not Windows; R Markdown and Quarto run selected code chunks rather
than rendering the document. Install jsonlite and rlang in the owning R environment. Parquet export also requires
nanoparquet 0.5.1 or newer. These workflows remain partial.
Use the same draft-and-code workflow with an R notebook dataframe.
DuckDB file editing is experimental and partial. DuckDB notebook relations and PySpark notebook dataframes are view-only. PySpark uses an existing local Classic or Connect batch session; Open Wrangler does not install or configure Spark. Streaming dataframes and remote or authenticated clusters are unsupported.
See feature parity and current limits, accessibility and keyboard use, the product gallery, and the dated Data Wrangler comparison.
Open Wrangler is an open-source project inspired by Microsoft Data Wrangler. It was built independently, uses no Microsoft Data Wrangler code or assets, and is not affiliated with Microsoft.
Contributions are welcome. See CONTRIBUTING.md, use GitHub Issues for bugs and feature requests, and follow SECURITY.md for vulnerability reports.
Open Wrangler is licensed under the MIT License.