lakeFS, the control plane for AI-ready data, is offering lakeFS for Agentic AI, briging governed, reproducible data access to autonomous and headless agentic workloads operating at enterprise scale.
According to lakeFS, agentic workloads make the data-readiness challenge much worse. Agents operate in parallel at machine speed across structured tables, unstructured files, images, video, and metadata, exposing the limits of manual governance and operational controls built for human-driven workflows.
lakeFS for Agentic AI addresses this directly. It gives every agent its own isolated data sandbox with a zero-copy branch of relevant data, validates and merges changes under policy, and produces a unified audit trail across every agent action, the company said.
"Agents are let loose on enterprise data at massive scale, but any agent that reads or writes to production data without isolation or a reproducible trail is a liability, no matter how good the model is,” said Einat Orr, CEO and co-founder of lakeFS. “The companies that win with agentic AI will solve this at the data layer and treat agents like production workloads, not experiments. That is what lakeFS has always done, and lakeFS for Agentic AI makes it explicit: the same control plane our customers use today is the one their agents need."
The core lakeFS technology that AI and data teams have been using at scale directly applies to autonomous agent workflows. lakeFS for Agentic AI is powered by its unique data version control architecture that provides zero-copy data sandboxing. It is built around the four pillars enterprises require before letting agents operate on production data, the company said. These pillars include:
- Isolation. Every agent works on its own zero-copy data branch, covering structured tables, unstructured files, and metadata together as one. Agent mistakes are automatically isolated and never corrupt production data.
- Reproducibility. Every agent run is tied to an exact, immutable version of the data. Past actions can be recreated, debugged, audited, or extended using the same inputs.
- Governance and compliance by design. Production data is gated by policy. Merges into production happen only after pre-merge validations pass. Every change can carry agent identity, run ID, and execution context. The result is a unified audit trail instead of evidence scattered across orchestrators, model providers, and cloud logs.
- Agent-native infrastructure. lakeFS provides file-level data access with branch-scoped credentials that confine each agent to its own workspace. That keeps each agent's working set narrow and avoids context bloat. No custom MCP server, SDK, or specialized integration is required.
lakeFS for Agentic AI is available now to all lakeFS Enterprise customers.
For more information about this news, visit https://lakefs.io.