CodeGraph takes a GitHub repository URL and turns its codebase into an interactive visual graph. It parses the source code, traces function calls, and explicitly maps out the relationships between different parts of a project so you can understand how everything fits together without having to manually grep through files.
- Interactive Node Graph: View the entire codebase as a network where every function is a node and every function call is an edge. You can click, drag, and play around with nodes to explore the architecture.
- Directory and File Views: Browse the repository structure in the sidebar. You can filter the graph to isolate the flow of a single file, or narrow down the graph to focus on a specific directory.
- Function Inspection: Click on any node to instantly see what file it lives in, its line numbers, and its direct dependencies.
- Semantic Chat: Ask plain language questions about the codebase (like "where is the billing logic handled?") and get direct answers backed by the underlying graph data.
Python, JavaScript, TypeScript, Go, Rust, Java, C, C++, Kotlin.
You need Docker and Docker Compose.
- Copy
backend/.env.localtobackend/.env - Add your Gemini API key (free at https://aistudio.google.com/app/apikey)
- Run
docker compose up --build
When you submit a repository, the system silently clones it, extracts the functions, builds the graph, and generates embeddings for search. Small repositories usually take under a minute.
You can jump between different repositories you've loaded using the dropdown menu. If a repository has been updated upstream, simply click the re-analyze button to pull the latest changes and rebuild the graph from scratch.