BLOG
Guides on auto-generated documentation, MCP tools for AI agents, git intelligence, and how repowise compares to other tools.
The same C4 container diagram written in Structurizr DSL, Mermaid and C4-PlantUML, with a feature table and an honest look at how fast each one goes stale.
SonarQube, CodeScene, Qlty, Codacy, NDepend and repowise compared on what their technical debt score measures and what evidence links it to bugs.
Tested commands to find circular dependencies in a monorepo: madge and dependency-cruiser for TypeScript, import-linter and pylint for Python, and Go.
Three ways to turn a GitHub repo into an architecture diagram, checked against Flask's real import graph: what a model draws well and what it leaves out.
Five ways to generate C4 diagrams from source code, which C4 levels each can derive, and how to regenerate them on every push so they stop going stale.
A day-by-day checklist for learning a new codebase in your first week, with a done state for each day and questions to ask the repo before your team.
Generate a JS or TS dependency graph with madge, dependency-cruiser, Nx, esbuild or repowise, with runnable commands and the output each one gives.
Codebase intelligence, or repository intelligence, is a kept-current model of a repo: code graph, git history, docs and health, served over MCP.
Claude Code searches with grep and keeps no index. Six MCP servers that add one, with measured tokens, tool calls and setup commands for each.
Built-in GitHub connectors make you pick files and hit token limits. How to give Claude.ai or ChatGPT a whole private repo through a remote MCP server.
Context7 serves library docs to coding agents. The alternatives, checked in October 2026, and how to give agents docs for your own codebase over MCP.
A neutral matrix of seven context tools for coding agents (Augment, Sourcegraph, Greptile, CodeGraph, Serena, GitNexus, repowise) and how to test them.