NVIDIA Config Manager

Overview

NVIDIA Config Manager is a network operations platform for bootstrapping and maintaining NVIDIA AI-factory infrastructure. It best suits network administrators and operators already prepared to deploy software on Kubernetes. Its broad workflow coverage and open-source license are appealing, but Developer Preview status makes it unsuitable for production use today.

Key features

Config Manager brings provisioning and recurring maintenance into a single workflow system. Zero-touch provisioning, device inventory, configuration backup and deployment, and reprovisioning cover equipment setup and lifecycle work. Operators can also validate cabling and hardware, rotate passwords, and upgrade operating systems or firmware. That range can consolidate routine work for teams managing NVIDIA AI-factory networks, though it does not eliminate the operational burden of running the platform itself.

Workflow orchestration and template rendering support repeatable operations. Configuration is stored in PostgreSQL, and the platform generates Kea DHCP configuration. Nautobot is the default DCIM provider; teams using another provider can integrate it through the provider-neutral DCIM SDK. Browser authentication uses an OIDC proxy, while API clients use bearer tokens. Approval workflows and an audit trail for configuration changes provide useful governance for day-two operations.

Installation uses Helm on a Kubernetes cluster, with either an interactive TUI or headless CLI installer. Cloud-hosted and on-premises Kubernetes deployments are documented. For offline sites, an airgapped bundle can include the chart, container images, dependency charts and manifests, and installer package. The standard requirements—Kubernetes access, kubectl, Helm 3.x, Python 3.13 or later, and uv—make this a better fit for infrastructure teams with Kubernetes experience than for organizations seeking a turnkey network appliance. Some workflows also use Docker and Kind. The large production or high-availability validation profile calls for 96 total vCPUs, 256 GB RAM, and at least three nodes, a substantial footprint even before considering the preview warning.

Pricing

Config Manager is open source under the Apache 2.0 license. Its free plan costs nothing and is marked Developer Preview, not recommended for production use. There is no free trial, and the stated offer is a free plan rather than a time-limited evaluation. This is a strong fit for teams assessing the software or building a non-production deployment; the preview status is the key trade-off, not a paid tier with reduced capabilities.

Platforms

Config Manager is listed for API, Linux, self-hosted, and web use. In practice, deployment is centered on Kubernetes with Helm rather than a conventional desktop installation.

Who it's for

The primary users are network infrastructure administrators who deploy and maintain the platform, and operators who run workflows and manage network operations. It is most relevant to teams responsible for NVIDIA AI-factory networking that can support Kubernetes and accept preview software in non-production settings. Organizations requiring production-ready deployment should look elsewhere for now.

Pros and cons

• Broad network workflow coverage: provisioning, validation, configuration backup, password rotation, reprovisioning, and OS or firmware upgrades address both setup and ongoing operations.
• Open, adaptable deployment: Apache 2.0 licensing, cloud or on-premises Kubernetes options, and an airgapped bundle suit teams needing control over hosting and connectivity.
• Integration and change controls: the DCIM SDK supports providers beyond default Nautobot, while approvals and an audit trail help govern configuration changes.
• Substantial operational prerequisites: Helm-based Kubernetes deployment and the large validation profile demand infrastructure expertise and resources.
• Preview limitation: the repository does not recommend production use, which rules it out for teams needing a production-ready network operations platform.

Alternatives

For broader network-provisioning options, browse Network Provisioning Software.

• PacketFence is worth comparing if you want a freemium option with a free trial; its self-hosted GPL v2+ edition offers unlimited devices and full source code.
• UniFi Network suits readers seeking free network software with a license-free Site Manager and a Network Server that can be self-hosted or run on UniFi hardware.
• TP-Link Omada SDN is a fit for centralized management of up to 1,500 devices at no additional expense, with capacity dependent on the PC or server hardware.
• Arista NG Firewall is a free alternative for readers focused on basic security and connectivity features with limited capabilities.
• Cambium cnMaestro offers a free Essentials tier for up to 10,000 devices and 10 management users, and a free trial.
• RUCKUS Cloudpath Enrollment System is a paid option with user-based subscription tiers and support included.
• Infoblox Cloud Network Automation uses token-based licensing, with tokens purchased as an allowance over an agreement term.
• EfficientIP SOLIDserver DDI is another paid option offered through hardware and virtual appliances.

Verdict

Choose NVIDIA Config Manager if your team manages NVIDIA AI-factory network infrastructure, wants open-source workflow automation, and can evaluate it outside production on Kubernetes. Its mix of provisioning, maintenance, integrations, and change controls is the main reason to consider it. Look elsewhere if production readiness is required: the Developer Preview warning is decisive.