Show HN: I built a codebase analysis MCP: Stria (github.com)

Stria, a newly developed structural codebase indexer, is transforming how LLM agents interact with code repositories by enabling sub-millisecond queries without the need for traditional parsing tools. Unlike conventional methods that rely on extensive abstract syntax tree (AST) parsers, consuming significant RAM and time, Stria employs a grammar-free phrase extraction approach. This innovation allows it to index large codebases, such as the 3.1GB Linux kernel containing 72,000 files, in just 80 seconds, with query execution times averaging mere milliseconds. The significance of Stria lies in its ability to provide LLM agents with a compact, efficient understanding of code structure. By generating a 50-token structural map for files and facilitating lazy loading of full function bodies only when necessary, it dramatically reduces the token costs typically associated with graph-based methods. Stria's design also removes the need for language-specific configurations, allowing it to seamlessly support any programming language and legacy systems. With minimal setup requirements and impressive performance metrics, Stria stands to enhance the efficiency of LLM tools in navigating complex codebases, making it a pivotal advancement for developers and the AI/ML community.