Code MRI App

An AI-powered repository analyzer that scans GitHub projects, builds code architecture graphs, and generates natural-language explanations to accelerate developer onboarding.

Comment

Inspiration

Understanding a new codebase is one of the hardest problems developers face. When joining a new project or exploring an open-source repository, developers often spend hours navigating folders, reading files, and trying to understand the architecture.

We wanted to solve this problem by building a tool that can instantly analyze a GitHub repository and explain its structure using AI. The idea behind Code MRI is similar to how an MRI scan reveals the internal structure of the human body — our tool reveals the internal architecture of a codebase.

Our goal was to help developers quickly visualize, understand, and explore complex repositories without manually reading thousands of lines of code.

What it does

Code MRI is an AI-powered repository analyzer that scans GitHub projects and generates insights about the codebase.

Users simply provide a GitHub repository link, and the system:

  • Crawls and analyzes the repository structure
  • Builds a visual architecture graph
  • Identifies relationships between files and modules
  • Generates AI explanations of the codebase
  • Helps developers quickly understand project structure

This dramatically reduces the time required to understand unfamiliar codebases.

How we built it

The system consists of a modern full-stack architecture:

Frontend

  • Built using React
  • Provides an interactive interface to submit repository URLs
  • Displays architecture graphs and AI-generated explanations

Backend

  • Built using Python with FastAPI
  • Handles repository crawling and analysis
  • Extracts file structure and dependencies

AI Layer

  • Powered by Gemini AI
  • Generates natural language explanations of the repository architecture

Infrastructure

  • Backend deployed on Google Cloud Run
  • Frontend deployed on Vercel
  • GitHub API used to fetch repository information

The backend processes repository data and sends structured information to the AI model, which generates developer-friendly explanations.

Challenges we ran into

One of the biggest challenges was accurately analyzing repository structures.

Some repositories have:

  • Deep folder hierarchies
  • Large numbers of files
  • Complex dependency relationships

Another challenge was handling GitHub API limitations and fallbacks. In some cases the API failed, so we implemented a crawler fallback system to ensure the repository could still be analyzed.

We also faced challenges in visualizing architecture graphs in a way that remained readable even for large projects.

Accomplishments that we're proud of

  • Successfully building a working AI-powered repository analyzer
  • Deploying a full-stack system with cloud infrastructure
  • Generating architecture graphs from real repositories
  • Integrating AI explanations to simplify complex codebases

Most importantly, we built a tool that can save developers significant time when exploring unfamiliar codebases.

What we learned

During this project we learned:

  • How to design AI-assisted developer tools
  • How to combine code analysis with generative AI
  • How to build and deploy cloud-native applications
  • How to visualize complex system architectures

We also gained deeper insights into how repositories are structured and how AI can help developers understand them faster.

What's next for Code MRI

We plan to expand Code MRI with:

  • Deeper dependency analysis
  • Support for multiple programming languages
  • Security and vulnerability insights
  • Better architecture visualization
  • IDE integrations (VS Code extension)

Our long-term vision is to build a platform where developers can instantly understand any codebase using AI.

Built With

Share this project:

Updates

posted an update —

Frontend Improvements Update

Shipped a set of UI/UX and responsiveness enhancements to improve usability across devices:

  • Fixed layout clustering and made the UI fully responsive (desktop → mobile)
  • Eliminated horizontal scrolling and improved layout stability
  • Refined graph view into a clean, controlled canvas interface
  • Added minimizable AI insights panel with smart positioning (no overlap with core UI)
  • Improved UX with better buttons, hover states, and interactive AI orb
  • Enhanced panel usability with sticky header and persistent controls

Result

A cleaner, more intuitive, and production-ready user experience with smoother interactions and better visual clarity.

Log in or sign up for Devpost to join the conversation.