Arcaide tackles program comprehension by generating multi-level call graphs that roll up function-level execution into classes and packages. Unlike traditional call graphs that explode exponentially, this approach lets developers zoom between abstraction layers — similar to C4 diagrams — expanding only the regions they need while collapsing the rest. The tool targets the growing challenge of navigating codebases increasingly written by AI agents rather than humans.

The graph feeds into an LLM for semantic analysis, which strips out telemetry and trivial utilities to further condense the view. It also identifies external interfaces and dependencies, enriching the structural map with behavioral context. The result merges a package diagram, class diagram, and use case diagram into a single composite view showing internal composition, external services, databases, and user interactions.

The creator positions Arcaide as a navigational aid, not a replacement for reading source. As development shifts from line-by-line coding to orchestrating code-generating agents, maintaining architectural awareness becomes critical. The tool runs on arcaide.foo with example diagrams available for feedback.