GitHub

CodeBoarding is an AI codebase documentation tool that generates interactive architecture diagrams by combining static control-flow analysis with a small LLM layer that names the clusters. Built for developers and coding agents, it maps the real architecture of a project instead of just listing folders or asking an LLM to index everything. It produces high-level system diagrams plus component-level documentation at multiple depth levels, and exports Mermaid-formatted output you can embed directly in docs and pull requests. Incremental analysis lets it focus on the parts of the code that changed, so reviewers can see what an AI agent built before it breaks. CodeBoarding ships as a CLI, a VS Code extension, and a GitHub Action, supports eight languages (Python, TypeScript, JavaScript, Java, Go, PHP, Rust, and C#), and works with multiple LLM providers including OpenAI, Anthropic, Google, AWS Bedrock, and Ollama. The project is open-source under the MIT license, with local CLI usage and a hosted diagram explorer at codeboarding.org. It is aimed at teams using coding agents who need readable, navigable architecture documentation that stays in sync with the code.

Developers use CodeBoarding to auto-generate architecture diagrams and documentation that map a codebase's real structure.