Using codebase-memory-mcp to find dead code in a Django project (49K nodes)

[DeusData]

Indexed a full Django project (fastapi-clean architecture, 335 nodes / 859 edges) and ran dead code detection:

search_graph(label="Function", relationship="CALLS", direction="inbound", max_degree=0, exclude_entry_points=true)

Found 12 functions with zero callers that aren't entry points (not route handlers, not main()). A few were genuinely dead — leftover utility functions from a refactor. Others were test helpers that only get called via pytest fixtures (which the graph doesn't track yet).

The key insight: exclude_entry_points=true filters out route handlers and framework-decorated functions automatically, so you don't get flooded with false positives. Without that flag, every route handler shows up as "dead code" because nothing in your codebase calls it directly — the framework does.

For larger codebases, you can scope it: search_graph(..., project="my-project") to avoid cross-project contamination if you have multiple repos indexed.

Anyone else finding interesting results with dead code detection? Curious what the false positive rate looks like in different frameworks.