Infrastructure teams are increasingly using agents to help plan changes, answer dependency questions, and keep infrastructure aligned with intent.
But that only works if the underlying data is connected, trustworthy, and rich enough for both engineers and agents to reason about relationships.
Infrahub has always stored your data in a graph, where the relationships between objects are first-class. With Graph Traversal in Infrahub 1.10, those relationships become something you can navigate directly, whether you’re investigating an issue yourself or using an agent as part of the workflow.
You can trace the path between any two objects, or find everything a given object depends on, from the UI, the GraphQL API, the Python SDK, or an AI assistant connected through the MCP server.
Built for the Way Infrastructure Connects
Real infrastructure is a web of relationships. A device connects to interfaces through bays and line cards; an interface connects through cables and circuits to a remote device; a service depends on a chain of objects several hops deep.
Now, you can ask Infrahub to show you exactly how systems in your network connect, and what depends on them – visually, and as a backend API for engineers and agents to use.
Graph Traversal, our newest capability, is what makes this possible. It has two modes:
1. Path Traversal
Which answers the question “How are these two things connected?”
Pick a source and a destination, and Infrahub returns every path between them through the graph, with hop counts showing the length of each path.
Click a path to highlight it, then click on any object to jump to its detail page.
Path Traversal is an easy way to trace connectivity when you’re investigating packet loss, validating a design, or trying to understand how two systems are related.
2. Dependency Mode
Which answers the question “What’s affected if this fails?”
Pick a source object and the kind of object you care about (such as services, customers, devices), and Infrahub will discover everything that depends on it, up to a configurable depth.
You can use Dependency Mode to analyze blast radius and impact when a maintenance window lands or a device drops off the network.
Both modes use the same visualization Infrahub already uses for its schema, so the experience is consistent across the platform.
Watch a quick Graph Traversal demo →
A Foundation for AI-Driven Operations
The same traversal that helps an engineer is what an agent needs to operate safely.
Connected through Infrahub’s MCP server, an AI assistant can call Graph Traversal to determine how objects relate and what depends on a given object. That gives the agent a focused, relationship-aware model to work from (which, otherwise, it would’ve had to reconstruct itself).
For example, say you receive a circuit maintenance notification and you want your agent to parse it.
The agent would:
- First call the traversal API to find every dependency of that circuit.
- Then produce an impact report inside Infrahub, against governed data.
- Finally, propose a change to the network, which goes through the same branching, review, and approval workflow as any change made by a person.
You could also use agents to troubleshoot. Here’s a quick scenario:
How to Use It
Simply:
- Set a source object
- Choose a mode (Path Traversal or Dependency mode)
- Pick a destination (or a destination kind)
Infrahub will render the result as an interactive graph you can explore.
For automation and scripts, both Path Traversal and Dependency modes are exposed through GraphQL and the Python SDK. Since the same endpoints power MCP-connected agents, traversal is available to your AI workflows out of the box.
Also, Graph Traversal always reflects the context you’re working in. So if you’re operating on a branch, results reflect the state of that branch rather than production, and results only ever include objects you have permission to see.
Why This Matters
Infrahub’s job is to be the authoritative, connected data layer for infrastructure automation, and increasingly, for the agents operating on top of it. That job depends on two things being true at once: (1) your data has to be modeled accurately, and (2) your data has to be reachable by everyone (and everything) that needs to act on it.
With Graph Traversal, the same dependency model your schema already describes becomes something everyone can use, engineers, automation, and agents alike.
For network and operations teams, that means impact analysis in seconds. And for the AI-driven operations you’re building toward, it means agents that can finally reason about your infrastructure – on a foundation you control.
