Red Hat adds support for agentic AI development

As use of agentic AI accelerates, Red Hat is hoping to position itself as the critical behind-the-scenes plumbing and connective fabric.

To this end, the company has unveiled new desktop and developer suite functions, skills bundles, and a rolling Linux release to help enterprises move beyond the experimental phase.

Announced at Red Hat Summit today, the new features and services are included in the latest release of Red Hat AI, with no additional usage charge. Tools are not metered and usage is not limited, Red Hat execs emphasized in a briefing.

“We’re helping developers accelerate and own their AI strategy with the same rigor they apply to their core IT applications,” said James Labocki, senior director for product management at Red Hat.

Meeting developers where they are

To build a standardized path of sorts, and help developers build and scale agents from their local desktops, Red Hat is making its Red Hat Desktop generally available and enhancing its Advanced Developer Suite.

The company now provides commercial support for the Red Hat build of Podman Desktop, an app for creating, managing, and deploying containers on Linux, macOS, and Windows. Another new capability in Red Hat Desktop is isolated AI agent sandboxing, which allows developers to build and test AI agents on their local hardware while preventing unwanted AI actions that might negatively impact the host operating system (OS).

The company has also expanded integrations in Red Hat OpenShift Dev Spaces. This secure, zero-configuration development environment now integrates with the Amazon Web Services (AWS) Kiro coding assistant (now in technical preview). The environment already integrates with the Claude command line interface (CLI), Microsoft Copilot, Cline, Continue, Roo, and others.

Developers can now connect their preferred tools to their cloud-based integrated development environment (IDE), and use frontier or private models.

Red Hat Desktop is built on Red Hat Hardened Images and Red Hat Trusted Libraries to enhance security. Hardened Images is a curated catalog of trusted, stripped-down, and micro-sized container images scanned for security and functionality. Trusted Libraries provide curated Python packages built on Open Source Security Foundation (OpenSSF) frameworks with a software bill of materials (SBOM) and cryptographic signatures to provide supply chain transparency and verifiability.

With these integrations, developers can access images and libraries from their laptops while also connecting to local or remote OpenShift clusters for unit testing.

“Taken together, this is Red Hat saying: Agentic development is the next productivity unlock, but the path there doesn’t require customers to re-platform onto a turnkey service stack,” said Devin Dickerson, Forrester principal analyst.

An enhanced developer suite

The latest version of the Red Hat Advanced Developer Suite introduces a trusted software factory. Now in preview, this standards-based continuous integration/continuous delivery (CI/CD) integration is based on accepted Cloud Native Computing Foundation (CNCF) best practices and Red Hat’s internal build processes. It can be used as-is or tweaked and replicated.

Both the expanded Advanced Developer Suite and Red Hat Desktop “extend governed, production-mirroring environments down to the laptop,” Dickerson said.

Further, built-in AI-driven exploit intelligence uses code reasoning and vulnerability analysis to isolate exploitable code paths and cross-check across broader vulnerabilities. This will allow developers to focus on the highest-impact fixes.

These enhancements reflect an increased focus on the local developer environment, which is now a security perimeter, said Shashi Bellamkonda, principal research director at Info-Tech Research Group.

“When an AI agent is pair-programming with you locally, the same governance controls that protect production need to extend to the laptop,” he said. “That is a different shift in architecture.”

Giving agents Red Hat-specific skills

Betting on an agentic future, Red Hat is introducing a dedicated skills repository to turn AI agents into “Red Hat superusers.”

Skills are essentially specialized knowledge bases that give AI agents step-by-step workflows in specific ecosystems, whether they’re scanning logs, analyzing code, or performing other actions. New agentic skill packs provide agents with best practices for specific Red Hat offerings, including OpenShift, OpenShift Virtualization, and Security Reliability Engineers.

“We often talk about models as the engine of AI, but in an enterprise context, a model without specific skills is like a high-performance vehicle without a steering wheel,” Matt Hicks, Red Hat president and CEO, wrote in a blog post. Based on Red Hat’s internal experiences, high-value work is not just about prompts. “It is in the craft of building the evaluations and frameworks that allows AI to operate with transparency and verifiable logic,” Hicks wrote.

Developers can begin with a core skill that understands the Red Hat ecosystem; as they progress, agents can recommend other related skills. Agents are also plugged into Model Context Protocol (MCP) servers, which connect them to external systems without requiring custom integrations.

“The AI skills bundle treats agent behavior as portable, versioned, inspectable software rather than vendor-locked prompts,” Dickerson said.

‘Instant on’ with Hummingbird Linux

New Linux features and changes are often subject to freezes (sometimes for up to several months), while their stability is tested. Red Hat is looking to bypass this issue with Fedora Hummingbird Linux, a free, rolling release service that supports anonymous, agent-driven pulls for instant deployment

“It’s a no-cost, free as in beer and free as in freedom, operating system,” Gunnar Hellekson, VP and GM for the Red Hat Enterprise Linux (RHEL) business unit, said during a briefing. “It will update very quickly. It’s delivered to the community as soon as there are updates in the upstream communities.”

Hosted within the Fedora Project community, Fedora Hummingbird Linux aligns with the “instant-on” expectations of the agentic era, operating as a fully autonomous software pipeline at “lights out” factory speed. The service is built on the same automated infrastructure and pipelines as Red Hat Hardened Images, meaning its languages, runtimes, databases, and tools are free of known CVEs and accompanied by SBOMs.

Red Hat plans to make Fedora Hummingbird Linux a default option across developer-focused cloud providers; for enterprises, it will be most useful in stage two proof-of-concept projects.

“Fedora Hummingbird gives developers a fast-moving, agent-forward OS for experimentation,” Dickerson said. “This is impactful for enterprises that need to move skills across environments and audit what agents are doing.”

Red Hat is betting that enterprises will need two simultaneous operating system strategies, added Info-Tech’s Bellamkonda: A “stable, slow-moving” foundation for production systems, and a “fast-moving track” that keeps pace with open source innovation.

“Most vendors make you choose,” he said. “Red Hat is trying to make both available under one subscription relationship.”

A ‘fundamentally different developer’

The announcements at Red Hat Summit can be best understood as a “deliberate counter-positioning to the hyperscaler ‘agentic enterprise’ pitch,” said Forrester’s Dickerson.

Whereas AWS, Google, and Microsoft are racing to make agent development a “turnkey service-layer experience for cloud-native greenfield workloads,” Red Hat is meeting customers where they live, he said. That is: In hybrid environments, private cloud footprints, regulated industries, and “long-lived” infrastructure.

Red Hat serves a “fundamentally different developer” than hyperscalers do, he said. Google Cloud and AWS are optimized for organizations looking to consume agentic AI as a managed service via Gemini Enterprise, Bedrock AgentCore, or Copilot Studio, “where the platform abstracts the complexity and the customer trades control for convenience.”

Red Hat’s customers, however, operate predominantly in private and hybrid cloud, often in regulated or sovereignty-sensitive industries, and have both the “appetite and the obligation” to manage complexity themselves, Dickerson said. The interesting question for them isn’t “which agent SaaS do we subscribe to?” It’s “how do we get the productivity gains of agentic development without surrendering the architectural control that’s core to how we operate?”

Red Hat is meeting its developers where they are on hybrid infrastructure, while at the same time closing the ease-of-use gap with agentic features “rather than asking customers to close the gap themselves by abandoning their architecture,” he said.

“It’s a quieter narrative than ‘operating system for the agentic enterprise,’” Dickerson said. And for Red Hat’s market segment, “it will likely be the more durable approach.”

This article originally appeared on InfoWorld.