NetBox 4.6 is GA: New Foundations for the System of Record Era
Custom Objects v0.5 and Branching v1.0.3 land in the NetBox ecosystem, alongside NetBox v4.6 — the building blocks for the safe, extensible, automation-ready system of record.
NetBox started as a way to document the network. Today it’s the system that automation reads from, that audits demand evidence from, and that a growing class of teams is starting to hand to AI agents. That changes what it has to be.
A passive catalog is a place you write things down. A system of record is something teams, pipelines, and increasingly, autonomous agents read from and write to with confidence. The job of the open source project is to make that foundation strong enough to hold all of that weight, and to keep doing it in the open, where the community that already runs NetBox can build on it.
That’s what today’s release is about. NetBox v4.6 is generally available, alongside major updates to two long-running source available plugins: NetBox Custom Objects v0.5 and NetBox Branching v1.0.3. Together they round out a foundation built for what teams are actually being asked to do with their network model.
Custom Objects v0.5: model everything you run, and how it connects
A key challenge in some NetBox deployments is where the built-in schema doesn’t quite fit what your team actually runs. A customer record at an internet exchange. An application’s dependency on the racks, switches, and clusters underneath it. A DHCP scope at a service provider. When the model can’t represent the thing, teams reach for spreadsheets, overloaded custom fields, or shadow tools, and the system of record stops being authoritative for the things automation actually depends on.
Custom Objects v0.5 closes that gap. It builds on earlier releases with a more complete pattern for extending the NetBox data model without writing code, including polymorphic one-to-one and one-to-many relationships across object types. You define new object types in the GUI, the REST API, or via version-controlled YAML, and you wire up the relationships between them and the rest of NetBox. From a service, see every component it runs on. From an interface, see every service that depends on it. The objects you create behave like first-class NetBox objects: filterable, queryable, bulk-editable, scriptable, exportable, and ready for the REST API.
Plugins still have their place for complex extensions, and the plugin ecosystem continues to power some of NetBox’s most powerful integrations. Custom Objects covers a meaningful share of the cases that previously required either a plugin or a workaround, and it does so in a way the rest of the ecosystem can rely on.
Branching v1.0.3: the trust primitive for change
Branching has been one of the most ambitious capabilities in the NetBox ecosystem since it landed as a community plugin. The pattern resonated immediately: stage a change in a separate workspace, see exactly what it will do, then merge atomically or not at all. With v1.0.3, every part of that workflow is production-hardened.
Branching is more stable than ever. When reconciliation is needed, merge reports give structured guidance for resolving it. Test coverage has been substantially expanded. The branch you create is the branch that merges.
That sounds like an engineering improvement, and it is, but it’s also what makes Branching usable as a primitive that other things can be built on. Practitioners get safe bulk changes for IP renumbers, VLAN cutovers, equipment refreshes, site migrations, and mass decommissions. CI/CD pipelines get branch state in every webhook, so external validators can check changes before a human reviews. Compliance teams get a reviewable diff, a named author, and a timestamp on every merged change, the kind of evidence auditors increasingly insist on. And teams piloting AI agents finally have a way to let an agent propose a change without risking the canonical record: the agent works in a branch, a human reviews the diff, and only then does it merge.
NetBox v4.6: the project keeps advancing
NetBox v4.6 moves the project forward on three fronts:
- Modeling primitives that match how infrastructure actually deploys. Cable Bundles for grouped physical cable runs, Rack Groups for row/aisle organization independent of the location hierarchy, Virtual Machine Types for VM categorization, and VMs assignable directly to devices without an intervening cluster.
- Audit-grade change records. Clearer diffs for custom-field data, retained create-and-update records for stable user attribution, and changelog messages on bulk component creation. The foundation SOC 2, PCI, and CAB processes increasingly demand.
- A more mature foundation. REST API improvements (ETag-based concurrency control, cursor-based pagination, partial tag assignment, broader export-template context), new plugin extension mechanisms (declarative layouts, reusable UI components, plugin-registered model actions), a Django 6.0 upgrade, and performance work across cable-path tracing and bulk export.
Three threads, one direction: making v4.6 trustable for everything that reads from and writes to it.
NetBox crossed 20,000 GitHub stars in March, with 391 contributors, more than 14,000 commits, and is trusted by more than 10,000 teams worldwide. Both Custom Objects and Branching ship as source available plugins, free for all, forever, and pre-installed in every commercial NetBox subscription. We’ll keep building in the open, alongside the community, on the project the market already runs.
What’s next
Branching is a primitive. The interesting question is what teams build on top of it: workflows for review, approval, and policy enforcement. We’ll have more to say on that soon.
For now: NetBox v4.6 is generally available. Custom Objects v0.5 and Branching v1.0.3 are live. If you’re already running NetBox, upgrade, try them out, and let us know what you think!
