A dependency graph is a map of which components, packages or services rely on which others. Each item is a node and each reliance is a link, so you can trace what a given piece of software needs and what needs it. Build tools, package managers and security tools all use such graphs.

In a large codebase, the graph lets tools work out what a change affects. Instead of rebuilding and testing everything, a build system can rebuild only the components that depend on the changed code, which saves time. In security work, the graph shows which applications include a vulnerable library, directly or through another package.

A graph is only as accurate as the information used to build it. Dependencies that are loaded dynamically, declared outside the usual files or added at runtime may not appear, and a graph can include dependencies that are never actually used. Circular dependencies make ordering and impact analysis harder.

To get value from one, generate it automatically from build and package metadata rather than maintaining it by hand, check it against how the software really behaves, and use it to find tightly coupled areas that are hard to change safely.