What tools are used for data provenance?

In data engineering

Catalogue and governance platforms such as Collibra, and the lineage features built into Databricks and Snowflake. These are strong at tracing transformations inside a pipeline and are usually bought by a data platform team rather than by a lab.

In the laboratory

Electronic lab notebooks and LIMS platforms record who ran what and when, and the regulated end of the market adds audit trail modules built for 21 CFR Part 11 and GxP work. These are built around instruments and protocols, and are typically validated, expensive and slow to introduce.

The gap between them

Neither family was designed for the case that is now common: a small team, external collaborators, and automated agents producing results overnight. A pipeline tool has no concept of an experiment; a notebook has no concept of an agent with its own identity.

Molecule sits in that gap. Every file version carries a content hash, a timestamp, a version number and the member who changed it, an agent can hold its own membership with its own role and expiry date, and the whole record is readable over an API so it can live inside a team's existing stack rather than only in ours.