srinivasan-sundaresan95/orihime
🔌 MCP Serversrinivasan-sundaresan95
A cross-repository code knowledge graph MCP server for deep static analysis without reading source files.
Orihime revolutionizes how LLMs interact with large codebases by shifting from raw source-reading to structured knowledge graph querying. It utilizes tree-sitter to parse code into a graph database (KuzuDB), allowing for complex queries that standard RAG systems struggle to handle. With over 30 specialized tools, Orihime supports multi-hop taint analysis, enabling security audits against OWASP, CWE, PCI, and STIG standards. It excels at identifying entry-point reachability, performance hotspots, and license compliance issues. By abstracting the code structure into a queryable graph, it allows AI agents to perform sophisticated architectural analysis and security reviews while maintaining high context efficiency. This makes it an essential tool for enterprise-grade code analysis where token limits and context window management are critical constraints.
💡Highlights
- ├─95% fewer tokens vs source-reading
- ├─Multi-hop taint analysis support
- └─KuzuDB-backed graph indexing
🎯For
- ├─Security Engineers
- ├─Software Architects
- └─AI DevOps Engineers