repowise-dev / repowise
Local codebase intelligence that pre-indexes your code so AI agents stop wasting tokens rediscovering what you already know.
星标趋势
AI 分析
项目摘要
Repowise is a codebase intelligence platform that indexes code, call graphs, git history, tests, and design decisions locally, then surfaces insights via MCP for AI coding agents and humans. It provides code health scores, dead code detection, auto-generated documentation, and git analytics without requiring API keys or LLM calls.
为什么值得关注
Repowise directly addresses the context window tax that AI coding agents pay by pre-indexing codebase knowledge locally, claiming 31.6% fewer agent tool calls and 2.3x more defect detection than CodeScene. Its call graph achieves 0.976-0.995 precision verified against compilers, and it can index 58,924 files on a single laptop — making it a practical bridge between static analysis and AI-assisted development.
优势
- Zero LLM calls required for core analysis — graph, risk, health, tests, and dead code detection are fully local
- MCP-native integration designed specifically for AI coding agents, reducing redundant context discovery
- Strong empirical benchmarks with published comparisons against 5 tools and compiler-verified call graph precision
- Highly active development with 62 releases in 6 months and 137 contributors
局限性
- AGPL-3.0 license may deter commercial adoption despite commercial licensing option
- No Docker support, which complicates deployment in containerized environments
- Relatively new project (created March 2026) with limited long-term stability track record
使用场景
- Reducing AI coding agent token costs by pre-indexing codebase context
- Identifying dead code and technical debt across large repositories
- Generating architectural documentation and design decision records automatically
- Prioritizing code review focus areas using health scores and risk analysis