The gist

Repowise is a codebase intelligence tool that solves the context problem for AI coding agents. It indexes a repository across four layers—dependency graph, git history, auto-generated documentation, and architectural decisions. This provides AI agents with a deep, historical understanding of the code, allowing them to answer questions about not just what the code contains, but why it was built that way, improving query efficiency and answer quality.

What it does

Indexes a codebase into four intelligence layers: dependency graph, git history, documentation, and decisions.
Exposes codebase intelligence to AI agents via seven task-oriented MCP tools.
Analyzes multi-repo workspaces to find cross-repository dependencies and co-changes.
Mines git history to identify code hotspots, ownership, and hidden coupling between files.
Generates a local web dashboard to browse dependency graphs, module health, and architectural decisions.
Keeps intelligence layers synchronized with code changes via git hooks, a file watcher, or webhooks.

How it works

Users install Repowise via pip and run repowise init in their project directory to perform an initial, deep analysis of the codebase. The tool then runs as a local server, exposing its intelligence layers to any MCP-compatible AI agent through a set of predefined tools. It also includes a web dashboard for exploring the indexed data. Repowise is open-source, and subsequent updates after commits are incremental and can be automated.

Best for

This tool is ideal for developers using AI coding agents on large, evolving, or multi-repository projects where understanding historical context and architectural intent is crucial for effective analysis and modification.

Watch out for

The initial indexing process for a new codebase can be time-intensive, taking around 25 minutes for a 3,000-file project.