AtlasCode

Problem & solution statement Developers joining an unfamiliar codebase — whether fixing a production bug, onboarding to a new team, or making their first open source contribution, waste hours just building a mental model of what exists and how it fits together. They need to answer questions like: What technologies does this project use? Which files are relevant to this issue? Where are the tests? How do I even get started? This discovery phase is slow, frustrating, and poorly supported by existing tools. CodeAtlas solves this by automating codebase understanding from end to end. Point it at any public GitHub repository and it clones the repo, scans every file, detects the technology stack with confidence scores and traceable evidence, extracts dependencies by ecosystem, and maps import relationships between components, all without executing a single line of the target code. The result is a structured knowledge model stored as plain JSON on disk. That knowledge model is surfaced in two ways. First, through a React dashboard with dedicated tabs for Overview, Technologies, Dependencies, Architecture (an interactive graph powered by @xyflow/react), Repository file tree, and Database schema. Beyond understanding, CodeAtlas includes a Guided Contribution Mode designed for developers making their first open source contribution. It fetches the project's open GitHub issues, classifies each one by type (bug, docs, feature, refactor) and scope (small, medium, large), and recommends approachable starting points. Once an issue is selected, the system generates a step-by-step contribution plan identifying the most relevant files and a suggested test command. It then creates an isolated git branch automatically, guides the developer through writing the fix, runs the test suite on demand with live pass/fail feedback, and finally produces a commit and a ready-to-use Pull Request draft in Markdown. The target users are junior developers and first-time contributors who need structure and confidence, as well as AI coding agents that need reliable, queryable facts about a codebase. CodeAtlas is unique in combining static analysis, and a human in the-loop contribution workflow in a single lightweight tool that requires no database, no authentication, and no code execution from the analyzed repository.