DeusData / codebase-memory-mcp
A blazing-fast, zero-dependency MCP server that builds persistent knowledge graphs from codebases, giving AI coding agents sub-millisecond structural query capabilities with 99% token savings.
星标趋势
AI 分析
项目摘要
A high-performance MCP server written in C that indexes codebases into a persistent knowledge graph using tree-sitter AST analysis and hybrid LSP semantic resolution. It enables AI coding agents to query code structure in sub-millisecond time with 99% fewer tokens than file-by-file exploration, supporting 158 languages and shipping as a single zero-dependency binary.
为什么值得关注
Exceptional performance (indexes Linux kernel in 3 minutes, sub-ms queries) combined with remarkable token efficiency (99% reduction) makes it a force multiplier for AI coding agents. The pure C implementation with zero dependencies, SLSA 3 compliance, and backing by peer-reviewed research (arXiv) set it apart from typical MCP servers.
优势
- Blazing fast performance with sub-ms queries and millisecond-scale indexing
- Zero-dependency single static binary for macOS, Linux, and Windows
- Hybrid LSP approach combining tree-sitter parsing with semantic type resolution for 10 languages
- Strong security posture with SLSA 3, OpenSSF Scorecard, and VirusTotal scanning
- Comprehensive test suite with 5604 passing tests and active CI
局限性
- No Docker support which may limit containerized deployment workflows
- No examples directory which could slow onboarding for new users
- Relatively new project (created Feb 2026) with 256 open issues indicating active but ongoing maturation
使用场景
- Enhancing AI coding agents (Claude Code, Cursor, Aider) with fast codebase understanding
- Building code navigation and exploration tools for large repositories
- Reducing token costs when AI agents need to understand code structure