Pravetz acts as a dedicated Micro-Agent Protocol (MCP) server that transforms raw code repositories into an intelligent, queryable SQLite database. By pre-indexing files, code chunks, symbols, call graphs, and documentation, it empowers AI assistants like Cursor to navigate large codebases efficiently, understand architecture, search relevant information, and coordinate multi-agent development workflows without the overhead of re-streaming entire repositories. This structured approach facilitates sophisticated code discovery, architectural reasoning, persistent knowledge management, and organized task execution for enhanced developer productivity.
主な機能
01Exposes tools for lexical and semantic code search and navigation.
02Facilitates multi-agent task orchestration with a shared task pipeline.
030 GitHub stars
04Pre-indexes codebases into a SQLite database with language-specific parsing.
05Builds and queries call graphs for architecture-level analysis.
06Provides a persistent store for external documentation and architecture diagrams.
ユースケース
01Coordinating multi-agent development workflows (e.g., Planner -> Executor -> Tester) through a shared task system.
02Assisting AI agents in understanding codebase architecture and planning changes based on call graphs and diagrams.
03Enabling AI assistants to quickly find relevant code, functions, and definitions within large repositories.