What Bauplan is (not)

Bauplan is a data platform built on Apache Iceberg with Git-like semantics. It has two parts:

1. A Git-for-data catalog. Branches, commits, merges and time travel over tables that live in your object storage.
2. A serverless engine. Query and transform your data in Python or SQL. Each run executes in an ephemeral sandbox in the cloud.

Principles

Open formats, your storage. Iceberg Tables, stored in your bucket and exposed through the Iceberg REST protocol. Any engine that speaks Iceberg REST can read them, so your data is never locked in.

Everything is code. Tables are produced by code, and each commit records the code that created it. Every change to the lake is a script you can review, version, and rerun.

Branches make changes safe. A data branch is a zero-copy view of the whole lake, and creating one is cheap. You iterate on production data in a branch, and production changes only when you merge.

Agents first. Everything in Bauplan is available through a Python SDK, a CLI and an MCP server, so coding agents can operate it as well as humans can.

What Bauplan does not do

Scheduling. Triggers, retries and backfills belong to your orchestrator, which calls Bauplan to create a branch, run, audit and merge.

Automatic refresh. A table changes only when a run, an import or a merge changes it.

Ingestion from sources. Bauplan imports files that are already in object storage. You can't upload a local file directly; copy it to object storage first, or land data there with a data integration tool.