RHEL 10.3 and 9.9 Bring Production-Ready AI Diagnostics

Red Hat on November 11, 2026, announced the availability of Red Hat Enterprise Linux (RHEL) 10.3 and 9.9, making AI-assisted diagnostics a supported production feature alongside new operating-system integrity controls, AI monitoring dashboards and developer tool updates.

The Model Context Protocol (MCP) server for RHEL has moved from Developer Preview to general availability. It connects compatible AI assistants to RHEL system data for diagnostics, with access governed by default rather than giving an assistant unrestricted access. The launch follows Red Hat’s November 9, 2026 announcement, which had described both versions as planned for general availability in the coming weeks.

AI Diagnostics Reach Production

The MCP server is packaged as an RPM in the RHEL extensions repository. It can be installed with goose, Red Hat’s open-source AI command-line agent, using standard dnf commands, placing the assistant connection within the operating system’s established package-management process.

Historical Performance Co-Pilot (PCP) data adds past performance events to the information available to those assistants. Red Hat says this helps identify root causes of intermittent issues without manual data correlation; administrators can investigate performance problems, failing services and configuration issues using live and historical signals, with changes applied through operational controls.

The command-line assistant itself has a separate deployment option with a different readiness level. Customers can run it in a self-managed configuration using customer-hosted inference, keeping model guidance and sensitive operational context inside secure, air-gapped or disconnected perimeters, but that option remains in Developer Preview. Red Hat’s hosted path remains available for connected estates.

Sealed Images Extend Integrity Checks

RHEL 10.3 also makes sealed images for image mode generally available. The feature uses customer secure-boot keys to verify operating-system content from firmware and the bootloader through the kernel and root filesystem, holding the system to its signed state. The root filesystem is read-only, and Red Hat says the kernel refuses tampered content at read time rather than serving it.

The mechanism is meant to prove that running content matches what was originally deployed, an operational problem across distributed fleets. Red Hat says the preverified foundation can detect unauthorized modifications and simplify compliance for financial, healthcare and government deployments, extending integrity verification from boot into runtime.

Developer Tools, AI Telemetry and Upgrades

The releases include GCC 16, LLVM 22, Go 1.27,.NET 11, Node.js 26 LTS, PHP 8.4 and Rust 1.97. New grafana-pcp AI dashboards combine GPU, CPU and network performance telemetry for monitoring AI and machine-learning workloads.

That gives operations teams two related uses for performance data: a consolidated view of the infrastructure running those workloads and historical context that the MCP server can supply to an AI assistant during troubleshooting.

Red Hat has also introduced validated, supported Leapp upgrade paths from RHEL 9.9 to 10.3 and from RHEL 8.10 to 9.9. These are in-place routes, addressing the lifecycle work involved in moving existing systems to the new releases rather than requiring every deployment to start from a fresh installation.

The result is a split production path for enterprise AI operations. MCP-based diagnostics and sealed images are generally available, while organizations that need to keep assistant inference inside a disconnected or regulated perimeter have only a Developer Preview option for that self-managed workflow.