Show HN: I built a CLI tool to map your codebase for LLMs (github.com)

codemap is a new open-source CLI that generates a compact, visually structured “brain map” of a codebase intended to give LLMs immediate architectural context without burning lots of tokens. It emits a dense, tree-like, pasteable block that highlights important files, flattens empty intermediate directories, clusters files and strips extensions to save vertical space, and color-codes entries by language. The tool automatically ignores noisy folders (.git, node_modules) and assets, flags the top five largest source files, and produces output designed for quick human and machine consumption. Install options include brew (JordanCoin/tap) or cloning the GitHub repo and running make install; usage is simply codemap or codemap /path/to/project. Licensed MIT. For the AI/ML community this matters because succinct, structured context is a frequent bottleneck when using LLMs for code understanding, generation, or refactoring: codemap reduces token cost while preserving high-value signals about project layout and big files that influence behavior. Technically it’s a preprocessing/serialization step suitable for prompt engineering, agent grounding, or inclusion in retrieval contexts; teams can embed the pasteable map into prompts or use it to seed vector store metadata. Caveats: aggressive flattening and noise-suppression trade off detail for brevity, so projects with nuanced directory semantics or polyglot monorepos may need customization or extended config.