Redpanda Launches Agentic Data Plane and Acquires Oxla

Redpanda announced the availability of its Agentic Data Plane on October 28, 2025, positioning the infrastructure as a unified control layer that mediates how AI agents interact with enterprise data systems. The platform addresses what CEO Alexander Gallego describes as a generational shift where "AI agents now define how work gets done" and organizations need infrastructure purpose-built for agent autonomy rather than retrofitted solutions. The announcement arrives alongside Redpanda's acquisition of Oxla, a distributed query engine that provides SQL capabilities for agents accessing streaming and historical data. The ADP combines four technical layers: a low-latency streaming backbone for event workflows, the newly acquired Iceberg-native query engine, over 300 enterprise connectors, and a governance framework that logs every agent interaction for audit and replay. Redpanda built remote Model Context Protocol implementations with fine-grained authentication, agent templates for common data sources like Git and Jira, and a declarative runtime that works across cloud, self-managed, and air-gapped deployments.

The platform specifically targets CIO concerns around access control and observability. Unlike traditional application failures where engineers can trace errors through code commits, agent decisions lack such transparency—a failed credit approval cannot be debugged through version control. The ADP records all prompts, context retrievals, and autonomous actions in immutable audit trails, while enforcing agent-bound tokens with narrow scopes and short lifespans. Strategic partners including Poolside and Ness Digital Engineering have committed to building on the infrastructure, citing requirements for "trust, security, and auditability without compromise" when deploying agents in regulated enterprise environments.

Early access to the full ADP stack, including Oxla integration, became available following the October announcement, with demonstrations showcased at Redpanda's Streamfest conference in early November 2025.

Decoded

This launch signals that enterprise AI infrastructure is splitting into a distinct category separate from traditional data platforms. While competitors bolt agent features onto existing architectures, Redpanda designed from agent requirements backward—governance first, connectivity second. The timing matters: as organizations move beyond experimentation to production agent deployments, the liability questions around who accessed what data and why become existential risks. The Oxla acquisition particularly reveals strategic thinking around query patterns unique to agents, which differ fundamentally from human analysts. Agents need to merge real-time streams with petabyte-scale historical context, execute complex joins across disparate sources, and do it under strict permission boundaries that traditional warehouse architectures weren't designed to enforce. Expect competitors to scramble building similar governance layers, but retrofitting security into existing platforms historically proves harder than designing for it from the start.