// 01 > CONTEXT
GitCompass reads commit history to surface high-churn hotspots, velocity spikes, contributor burnout signals, bus-factor risk, and onboarding guidance. It ships as a CLI and a visual dashboard, and runs entirely on your machine.
// 02 > THE_PROBLEM
The signals that predict delivery risk, churn, single-owner files, after-hours spikes, are buried in Git history that nobody has time to mine.
Existing analytics meant shipping your repo to a cloud service, an immediate non-starter for private code, and still answered fixed questions rather than the one you actually had.
Before
Risk signals buried in history · cloud upload required · fixed dashboards only
// 03 > APPROACH
I built a modular monorepo: an analytics core, a CLI, and a dashboard, with an AI query layer that answers questions in plain English, all running locally.
Local-first by design
Everything runs on your machine. No account, no cloud, no telemetry, so it works on private repos.
Plain-English querying
Ask the repo a question in natural language; multi-SDK support over the local analytics.
// 04 > THE_RESULT
Hotspots, burnout, and bus-factor risk surface from one local scan, queryable in plain English, with nothing leaving the machine.
Before
- History too costly to mine
- Cloud upload to analyze
- Pre-canned questions only
After
- 5 risk signals from one scan
- 100% local, no telemetry
- Natural-language queries
“Analytics shouldn't cost you your privacy. GitCompass never phones home, and you can just ask it what you want to know.”