CodeBoarding
CodeBoarding is an open-source codebase analysis tool that turns large codebases into interactive, high-level diagram representations—so both humans and soon agents can understand them.
Designed for onboarding, documentation, and system comprehension, CodeBoarding uses static analysis and LLM agents to map out your project’s structure, then turns that into diagrams you can embed in your docs, IDE, or CI pipeline.
Inspiration
When I started programming, I remember spending hours drawing up diagrams for codebases just to understand the repo. Years later - I still find diagrams the best way to get a mindmap before diving into new code.
So I built this (fully free) open-source project to generate those diagrams automatically. It summarizes repositories into simple, visual representations that you can explore at different levels of abstraction.
Hope this helps. Please try it out and let me know what you think.
Demo
Run CodeBoarding on your own repo:
🌐 www.codeboarding.org/demo
Source code:
📄 www.github.com/CodeBoarding/CodeBoarding
How it works
Control Flow Graph Extraction
It starts with static analysis—extracting modules and their relationships based on control flow, not just import graphs.Multi-level Abstraction via LLMs
CodeBoarding uses an LLM agent (provider-agnostic) to organize these into conceptual modules, abstractions, and summaries.Diagram Generation
Outputs are rendered as Mermaid.js diagrams—interactive and embeddable in docs, IDEs, and web interfaces.MCP Integration (in progress)
The generated representation can be served via a lightweight MCP server, allowing AI devtools (e.g., ClaudeCode, Cursor, VSCode agents) to query context-aware summaries and structure.
Next steps
Support more languages: Support for additional static-analysis friendly languages (e.g., Go, TypeScript, C++) is on the roadmap. Feedback on which to prioritize is very welcome!
MCP Server: An MCP server is under development to allow your dev agents to use the overview for improved codebase indexing.
Top comments (0)