We’re Amin and Parsa, and we’re excited to share DataKit, a fully in-browser data analysis platform that lets you work with large datasets directly in a browser tab, with no servers, no setup, and no data leaving your machine.
- GitHub: https://github.com/datakitpage/datakit
- Live demo: https://datakit.page
DataKit processes multi-gigabyte datasets (CSV, Parquet, JSON, Excel) entirely client-side using DuckDB compiled to WebAssembly. Your data stays local to your browser, and nothing is uploaded anywhere by default.
We were frustrated by having to choose between cloud tools that require uploading sensitive data and heavyweight local setups that are painful to install and maintain. We wanted something that just works in a browser tab, but still has real analytical power.
Some of its core features are:
- Client-side processing of large files (tested up to ~20GB) with no backend
- Full SQL interface powered by DuckDB-WASM
- Python notebooks via Pyodide for data science workflows
- Optional connections to remote sources (Postgres, MotherDuck, S3) via a proxy
- An AI assistant that only sees schemas and metadata — never raw data
We’ve been building DataKit as a side project over the past few months and would really love feedback on:
- Performance bottlenecks you run into
- Features you’d need for your workflows
- Thoughts on the all-client-side architecture vs hybrid approaches
— Amin & Parsa