lakeFS

lakeFS is a data version control platform ("Git for data lakes") from Treeverse that brings zero-copy branching, provenance, and unified access to enterprise data. Its June 2026 lakeFS for Agentic AI offering gives every AI agent an isolated, reproducible data sandbox with policy-gated merges and a unified audit trail, aimed at data and platform teams running autonomous and headless agents at scale.

lakeFS, built by Treeverse, is an open-source and enterprise data version control platform that applies Git-like concepts—branching, committing, and merging—to data lakes, positioning itself as 'the control plane for AI-ready data.' Its core capability is zero-copy branching, which lets teams create isolated, fully consistent snapshots of production data for testing and experimentation without duplicating it. In June 2026 the company introduced lakeFS for Agentic AI, which extends this architecture to autonomous and headless agent workloads: each agent gets its own isolated data sandbox on a zero-copy branch with branch-scoped credentials, reads and writes through standard file operations (no custom MCP server or SDK required), and can only merge into production after pre-merge validations—schema checks, data-quality tests, row-count comparisons, and custom business rules—pass as an atomic operation. Every change can carry agent identity, run ID, and execution context, producing a unified audit trail rather than evidence scattered across orchestrators, model providers, and cloud logs. lakeFS is used by organizations including Amazon, Netflix, Arm, Lockheed Martin, Volvo, NASA, Bosch, and the U.S. Department of Energy, and is available as open source, a managed cloud service, and a self-managed enterprise edition.

Category

Data & Analytics

Pricing

Freemium, Subscription, Contact for pricing

Target Market

Data Engineers, Platform Engineers, CIOs, AI/ML Engineers

• ✓
  Zero-copy branching

  Creates isolated, fully consistent snapshots of production data for agents or experiments without duplicating the underlying objects.

• ✓
  Agent-native data sandboxes

  Each agent works through standard file operations with branch-scoped credentials confining it to its own workspace—no custom MCP server or SDK needed.

• ✓
  Policy-gated atomic merges

  Changes reach production only after pre-merge validations (schema checks, data-quality tests, row counts, custom rules) pass, applied atomically.

• ✓
  Unified audit trail

  Every change carries agent identity, run ID, and execution context for a single governed record of what each agent did.

• •
  Governed autonomous agents

  Let headless agents read and write enterprise data safely, gating any production change behind automated validation.

• •
  Reproducible data experiments

  Branch production data to test pipelines or models in isolation and roll back cleanly if results are wrong.

• •
  Data lineage and compliance

  Track provenance and produce audit evidence for AI-driven data changes across the lake.

Open-source edition is free and self-managed; Cloud and Enterprise editions add managed hosting and advanced governance—pricing on request.

This page was written from 3 sources, 1 on domains other than lakefs.io.

1. 1.lakefs.io — lakefs.iovendor
2. 2.lakefs.io — trust but verify agent data accessvendor
3. 3.businesswire.com — lakeFS for Agentic AI Isolated Reproducible Enterprise Data