Repowise — Codebase Intelligence for AI and Humans

Introduction

Repowise is an open-source codebase intelligence tool that analyzes your repository to produce health scores, detect dead code, generate documentation, and surface architectural decisions. It works both as a standalone CLI and as an MCP server for AI coding agents.

What Repowise Does

• Computes code health scores across complexity, coupling, and test coverage
• Detects dead code, unused exports, and orphaned files
• Auto-generates documentation from code structure and patterns
• Provides git analytics including contributor patterns and churn hotspots
• Serves insights via MCP for integration with coding agents

Architecture Overview

Repowise is a Python application that combines static analysis with git history mining. The analyzer walks the AST of supported languages to build a dependency graph, then cross-references with git log data to identify churn hotspots and ownership patterns. Results are stored in a local SQLite database for fast querying. The MCP server exposes these insights to coding agents.

Self-Hosting & Configuration

• Install via pip: pip install repowise
• Run analysis with repowise analyze <path>
• Configure analysis rules in .repowise.yml
• Start the MCP server for agent integration
• Schedule periodic analysis via cron or CI

Key Features

• Code health scoring with actionable recommendations
• Dead code and unused dependency detection
• Git-based churn and ownership analytics
• Auto-generated documentation from code patterns
• MCP server for AI agent integration

Comparison with Similar Tools

• SonarQube — Enterprise code quality; Repowise is lighter with agent integration
• CodeClimate — Hosted analysis; Repowise is fully self-hosted and open source
• Semgrep — Pattern-based security scanning; Repowise focuses on health and architecture
• Fallow — TS/JS codebase intelligence; Repowise supports multiple languages

FAQ

Q: What languages are supported? A: Python, TypeScript, JavaScript, Go, and Rust. More languages are planned.

Q: How does MCP integration work? A: Run repowise serve --mcp to expose codebase insights as MCP tools that any compatible agent can query.

Q: Does it require a database? A: Repowise uses an embedded SQLite database stored in your project directory. No external database needed.

Q: Can I run it in CI? A: Yes, use repowise analyze --ci for non-interactive mode with exit codes based on health thresholds.

Sources

• https://github.com/repowise-dev/repowise