Hey devs - built a tool that helps understand large codebases using AI + knowledge graphs
TL;DR: Created Markar - it parses your entire repo into a knowledge graph, then runs
autonomous AI agents on top. Tells you exactly what breaks before you merge code,
generates meaningful tests automatically, and does code reviews that actually understand context.
Problem we solved:
- Large codebases are incomprehensible
- Code changes have unpredictable impact
- Code reviews miss cross-module issues
- New devs take months to be productive
How it works:
- Parse code → build knowledge graph (functions, classes, dependencies)
- Real-time updates as code changes
- AI agents reason about the graph to generate insights
Currently testing with 20+ companies. Seeing 40% reduction in production bugs and 2x faster releases.
Open to feedback and technical discussions.
Top comments (0)