Redpanda brings identity, policy control, and data governance to AI agents
Redpanda announced the availability of new core capabilities in the Redpanda Agentic Data Plane (ADP), including a centralized AI gateway, AI observability and evaluation via OpenTelemetry, AI agents, and unified authentication and authorization. Together, these features form a unified governance layer that allows enterprises to securely connect AI agents and Model Context Protocol (MCP) servers to live enterprise data with visibility and control.
As organizations move from AI experimentation to production, the challenge has shifted from building agents to governing them.
“This is why the enterprise agentic AI market is struggling while consumer AI flourishes,” said Tyler Akidau, CTO of Redpanda. “Building an agent is remarkably easy. Running one safely in a company, with access to sensitive systems and data, remains genuinely hard. Redpanda is focused on bringing control to the connectivity layer – rather than configuring access policies individually at each data source, enterprises need a central point through which all agent interactions flow: an agentic data plane.”
“AI agents don’t fail because models are bad; they fail because systems lack control,” said Alex Gallego, founder and CEO of Redpanda. “With Redpanda Agentic Data Plane, we’re giving enterprises a practical way to operate agentic systems safely, with identity, policy enforcement, and observability built in from day one.”
Available Redpanda ADP capabilities
First introduced in October, Redpanda’s ADP exists to make AI agents trustworthy at scale. Acting as both a governance layer and a control plane, it governs how agents authenticate, access data, and take action while recording intent, inputs, and outputs for complete auditability. With AI Gateway, open agent interoperability, unified identity, and full observability, enterprises gain the controls needed to safely run agents on live data without sacrificing speed or visibility.
“Redpanda addresses the right requirements with this release, as our research shows a strong need for both strong AI governance and real-time data inputs,” says Kevin Petrie, VP Research and Head of Data Management Practice of BARC. “AI adopters feel overwhelmed with the complex task of meeting these requirements while safely integrating powerful GenAI models into their business processes. Redpanda’s approach will help make data and AI leaders’ lives easier.”
AI Gateway
Redpanda AI Gateway provides a unified access layer between applications, AI models, and MCP services. It centralizes routing, policy enforcement, cost controls, and observability across all AI traffic. Enterprises can define token budgets, set spending limits, and optimize usage through deferred tool loading, bringing operational discipline to complex agent workflows.
The AI Gateway also serves as a central point to aggregate and govern MCP servers, with an admin-controlled registry of approved servers and YAML-based configuration for rapid deployment.
AI agents
Redpanda ADP is designed to work with any AI agent framework. Enterprises can run and govern agents built on their existing frameworks, as well as use Redpanda’s built-in AI agents when they want a fully-managed option. All agents, whether hosted by Redpanda or external, interact with data and tools through open standards like the A2A protocol, and integrate with ADP’s unified authentication, authorization, and observability services. Agents can also be triggered through Redpanda Connect pipelines, enabling real-time, event-driven and human-in-the-loop workflows.
Unified authentication and authorization
All components of the Redpanda Agentic Data Plane are secured through OIDC-based identity and fine-grained authorization policies. Every request, whether from a user, service account, or agent, is authenticated and governed, eliminating long-lived credentials and reducing the risk of uncontrolled agent access.
End-to-end observability and evaluation
Redpanda ADP emits full-fidelity metrics, traces, logs, and transcripts using the OpenTelemetry Protocol (OTLP). Enterprises can inspect agent behavior directly in the Redpanda console or export traces to external observability platforms, enabling debugging, compliance, and post-incident analysis.
The Redpanda Agentic Data Plane is built on a low-latency streaming foundation, enabling real-time data access, continuous context updates, and event-driven agent execution. With more than 300 connectors available through Redpanda Connect, enterprises can expose data from databases, SaaS platforms, data streams, and data lakes without moving data or breaking existing architectures.