lakeFS

lakeFS is an open-source, Git-style version control layer for data lakes that lets data engineering and ML teams branch, commit, merge and roll back petabyte-scale object storage without copying data. It suits organisations that need reproducible training datasets, isolated testing on production data and an audit trail for AI data governance. Enterprise and managed Cloud editions add RBAC, SSO and hosted operations.

lakeFS is an open-source data version control system from Treeverse that turns object storage such as Amazon S3, Azure Blob Storage, Google Cloud Storage and S3-compatible stores like MinIO into a Git-like repository, so teams can branch, commit, merge, tag and revert data the way developers manage code. Branches are zero-copy: lakeFS creates them without duplicating the underlying objects, which lets engineers test pipelines, run ML experiments or let AI agents work against a full production snapshot in isolation, then merge atomically or roll back a bad load. Applications reach it through an S3-compatible gateway, generated API clients and a high-level Python SDK, and hooks fired on commits, merges and branch creation enforce write-audit-publish quality gates. The core is written in Go under the Apache 2.0 licence; the GitHub repository showed about 5.5k stars, 479 forks and 359 open issues in September 2026, with five releases between 9 July and 5 August 2026 (latest v1.86.0). Treeverse, founded in 2020 by Einat Orr and Oz Katz and headquartered in New York, first released lakeFS in August 2020, launched the managed lakeFS Cloud in June 2022 and shipped 1.0 in October 2023. It now markets the product as a control plane for AI-ready data, adding lakeFS Mount, an Apache Iceberg REST catalog and transactional mirroring. The free Community edition runs with a single administrator user; RBAC, SSO and SCIM belong to lakeFS Enterprise, offered as a managed service on lakeFS Cloud or self-hosted on-premises, in a customer's own cloud or air-gapped. Treeverse raised a $20M round led by Maor Investments in July 2025, taking total funding to $43M, and acquired the DVC project in November 2025. Named users include Arm, Bosch, Lockheed Martin, NASA, Volvo and the U.S. Department of Energy.

Category

Data & Analytics

Pricing

Free, Contact for pricing

Target Market

CTOs, CIOs, Data Engineers, ML Engineers, Data Scientists

Deployment

Open-source, Self-hosted, Hybrid

Founded

2020

Headquarters

New York, United States

Customers

Thousands of organizations use lakeFS (company claim, July 2025)

• ✓
  Zero-copy branching

  Creates isolated branches of an entire repository without duplicating the underlying objects, so teams can test changes on full production data cheaply.

• ✓
  Atomic commits, merges and reverts

  Commits are immutable snapshots and merges update a branch atomically, which makes instant rollback after a bad pipeline run or data incident possible.

• ✓
  Hooks for write-audit-publish

  Hooks triggered on commits, merges and branch creation run validation checks before data reaches consumers, turning quality rules into enforced gates.

• ✓
  S3-compatible gateway

  Exposes repositories through an S3-compatible API, so Spark, Athena, Presto, DuckDB and boto3 clients work with minimal code changes.

• ✓
  Format-agnostic versioning

  Versions Delta Lake, Iceberg, Hudi and Parquet tables alongside unstructured images, video and sensor data under one versioning model for multimodal AI datasets.

• ✓
  Apache Iceberg REST catalog

  Adds an Iceberg REST catalog, announced in mid-2025, so Iceberg-compatible engines can work with version-controlled structured tables through the standard catalog protocol.

• ✓
  lakeFS Mount

  Mounts a lakeFS repository locally so ML training code and notebooks can read versioned data as ordinary files without custom loaders.

• ✓
  Enterprise access control and mirroring

  lakeFS Enterprise adds role-based access control, single sign-on, SCIM provisioning and transactional mirroring, which the single-admin Community edition lacks.

• •
  Reproducible ML training

  Tag the exact commit a model was trained on so auditors or engineers can reproduce the training dataset months later for compliance reviews.

• •
  Testing pipelines on production data

  Run a changed Spark or dbt pipeline on a zero-copy branch of production data, validate the output, then merge atomically instead of testing on stale samples.

• •
  Data incident rollback

  When a bad load corrupts a table or dataset, revert the branch to the last good commit instead of restoring from backups, cutting downtime for consumers.

• •
  Sandboxes for AI agents

  Give AI agents an isolated branch of real data to read and write, then review and merge their changes only after automated checks pass.

• •
  Write-audit-publish quality gates

  Use pre-merge hooks with tools such as Great Expectations to block data that fails quality checks from reaching the production branch.

The Community edition is free under the Apache 2.0 licence. lakeFS publishes no list prices: its pricing page describes only lakeFS Enterprise, delivered as a managed service on lakeFS Cloud or self-hosted on-premises, in a customer cloud or air-gapped, and routes buyers to sales. The metering basis for paid plans is not disclosed.

This page was written from 12 sources, 8 on domains other than lakefs.io.

1. 1.lakefs.io — lakefs.iovendor
2. 2.lakefs.io — pricingvendor
3. 3.lakefs.io — securityvendor
4. 4.docs.lakefs.io — access control lists
5. 5.docs.lakefs.io — python
6. 6.lakefs.io — celebration shared vision lakefs dvcvendor
7. 7.github.com — lakeFS
8. 8.github.com — releases
9. 9.hn.algolia.com — 41061786
10. 10.prnewswire.com — git for data pioneer lakefs secures 20m in growth capital fi
11. 11.dbta.com — lakeFS Secures 20M in Growth Capital Transforms Critical Gap
12. 12.en.wikipedia.org — LakeFS