CodeBoarding: Interactive Diagrams for Codebases

An open-source project, CodeBoarding, generates interactive diagrams for codebases to aid visual learning. It leverages static analysis and large language models (LLMs) for improved accuracy. The diagrams are free and accessible via GitHub. A key challenge was preventing LLM overload; this was addressed using control-flow graphs to focus prompts. Continuous integration/continuous deployment (CI/CD) is supported by prompting the LLM with Git diffs. Future development includes a model context protocol (MCP) server for IDE integration. The MCP will intelligently route queries to relevant code sections within LLM context window limitations. The project currently supports Python, with future language support under consideration. The creator seeks feedback on LLM hallucinations in large codebases and the utility of a lightweight MCP. This is the creator's first public project.