Architecture

Next.js 16 frontend with Framer Motion animations, talking to the Anthropic Claude API to analyze real git history scraped from 10 major open-source projects. Honest about what it analyses and what it doesn't.

Case study

AI Scrummaster

Team intelligence from real git history.

Most "AI for engineering teams" demos are hand-waved over fake data. AI Scrummaster is built on the actual commit history of 10 major open-source projects — Kubernetes, Rust, Django, PyTorch, Grafana, and others — and uses Claude to surface what's real about how teams actually ship software.

The honest scope: it reads git, it doesn't read minds. It tells you what the data says, then it gets out of the way.

What it does

• Reads real git history from 10 production OSS repositories
• Surfaces team health signals: velocity changes, commit cadence, contributor churn, patterns of blocked work
• Uses Claude for synthesis — not as a magic oracle, as a way to compress patterns into prose a Scrum Master can act on
• Polished frontend — Next.js 16, React 19, Framer Motion animations
• Honest documentation — explicit about what it can and cannot do

Status

31 commits, brand new (April 2026). Polished enough to demo. Built fast as a "what if a Scrum Master could actually see the data?" experiment.

Tech stack