Real-time data streaming company Redpanda Data Inc. said it’s looking to tackle the regulatory challenges involved in streaming information to artificial intelligence agents with the launch of its new agentic data plane.

The launch stems from Redpanda’s recent acquisition of the Structured Query Language engine startup Oxla sp. z o.o., which was also announced today. The Redpanda agentic data plane is meant to serve as a reliable data access layer for building AI agents that can be traced, observed and governed within enterprise environments.

San Francisco-based Redpanda is backed by more than $160 million in funding from Alphabet Inc.’s GV fund and other investors, and has created a namesake platform that’s used by enterprises to move data between applications in real time. For instance, a manufacturing firm might use the platform to stream malfunction alerts from a piece of factory equipment to a cloud-based monitoring tool.

The company claims it can move data between systems faster than rival tools such as the open-source Apache Kafka platform, partly because its tool is written in C++, a more performant programming language than Java.

Redpanda said its agentic data plane combines the capabilities of its battle-tested data stack and connectivity suite with Oxla’s standard SQL query engine and database tools. The Polish startup is the creator of a self-hosted data warehouse platform that’s designed for large-scale, low-latency analytics and optimized for compute, memory and storage efficiency, enabling companies to manage growing datasets with predictable costs.

The company seems to have acquired Oxla specifically to get its hands on the startup’s technology, which provides the governance tools for its agentic data plane. With Oxla, it can integrate ready-made standards for the open-source Model Context Protocol and SQL alongside the existing standards it supports, such as the Apache Iceberg data table format.

Redpanda co-founder and Chief Executive Alex Gallego said the agentic data plane is an essential piece of the puzzle for AI agent deployments. “AI agents have made initial inroads to the private networks of the modern enterprise,” he said. “But organizations don’t have an easy, integrated way to observe, control and govern data access for these agents.”

According to Gallego, AI agent deployments that lack an agentic data plane are likely noncompliant, because they’re being given permission to read, write and process enterprise data and share it with large language models that autonomously take actions based on the data they see. “The agentic data plane provides the very necessary MCP-aware connectivity, access control and governance required for enterprise AI,” Gallego said. “With the acquisition of Oxla, we can now give agents a standard SQL endpoint to query data in motion or at rest.”

Gallego said the agentic data plane enables unified connectivity across the enterprise, combining with its data movement platform to bridge private and public cloud environments. It offers full observability into what AI agents are doing, monitoring every single agent interaction, including the prompts, inputs, context retrieval and actions they take. In addition, it supports federated querying – where agents query multiple data sources at once – and data transformation, with its SQL query engine making it simple for agents to join and transform data across disparate sources in real time.

Eiso Kant, co-founder and co-CEO of the AI company Poolside Inc. said his company is using Redpanda’s agentic data plane to support its suite of intelligent agents that perform software coding assignments on behalf of human developers.  Nvidia Corp. plans to invest as much as $1 billion in the artificial intelligence company

“It provides us with a real-time data platform that gives agents the context they need to operate in highly secure enterprise environments,” Kant said. “Together, we’re closing the gap between code generation and live execution, moving organizations from AI experimentation to autonomous systems running in production.”

Image: SiliconANGLE/Dreamina AI