omertt27/codemap

CodeMap describes itself as Google Maps for codebases. Point it at any repository and it builds an interactive map of that codebase: files, folders, imports, exports, functions, classes, and how they all connect through a dependency graph. It runs entirely on your own machine, so no code ever leaves your computer. It currently understands Python, JavaScript, TypeScript, and Java. You install it globally with npm, or run it once without installing using npx, then point the codemap command at any repository path. The main command opens an interactive map in your browser, while other commands scan a repo and save its graph, generate governance reports, or analyze the architectural impact of a pull request. Everything the tool produces is written into a local .codemap folder inside the scanned project, which you would normally add to your gitignore file. Beyond the map itself, CodeMap works as an architecture guardian. It computes a health score from zero to one hundred with a letter grade, based on measurements like maintainability, stability, modularity, coupling, and complexity. These scores come from deterministic rules and graph algorithms rather than an AI model, so the same codebase always produces the same score, making it useful for tracking whether a project's structure is improving or decaying over time. You can configure custom rules, such as a maximum file size or forbidding one part of the codebase from importing another, in a configuration file at the repo root. Other commands let you see the blast radius of changing a specific file, compare architecture between two git revisions, review how the codebase evolved over its history, and export the whole graph as structured JSON for use by other tools. It can also run as a local MCP server, exposing all of this analysis directly to AI coding agents so they can query the codebase's structure while working on it. Running it requires Node.js version 22 or later, and git is needed only for the history and time machine features.