Progress Reports Die When Every Git Commit Updates the Kanban Board
Progress tracking that depends on memory or separate tools consistently drifts from reality. Hooking into `git commit` — the one action every developer already performs — eliminates the reporting gap without adding a new workflow, and the offline queue means it works even when the network doesn't.
MaDu is a lightweight progress-tracking tool built around a single idea: commit code, and the kanban updates itself. A CLI hook intercepts `git commit` and prompts the developer to select a task, bug, and progress percentage from a numbered list. The commit message gets formatted with a task ID and progress marker, and the data syncs to a web dashboard. Offline commits queue locally and upload when connectivity returns.
The web platform shows projects with battery-style progress bars (red below 30%, blue 30–79%, green 80%+), member commit logs, and auto-generated daily, weekly, and monthly reports. AI features handle three friction points: pasting a PRD or uploading a .docx auto-splits it into structured tasks; `git diff` analysis suggests a commit message; and a one-click report generator turns commit records into prose summaries.
A quick-command mode lets experienced users bypass the interactive prompts entirely — `madu done --task TASK-1001 --progress 70 --type task-iteration -m "message"` does the commit and the reporting in one line. The tool targets teams of 5–20 developers who find Jira too heavy and manual tracking too unreliable.
The core insight is that progress reporting fails because it's a separate act from the work itself; attaching it to `git commit` makes it a side effect rather than a chore.
The offline queue design acknowledges a real-world condition — developers commit on trains, planes, and in coffee shops — that most SaaS project tools ignore.
MaDu explicitly positions itself between "too heavy" and "nothing," targeting the 5–20 person team gap where Jira is overkill but a spreadsheet is chaos.
The AI features are applied to three narrow, high-friction tasks (PRD breakdown, commit message writing, report generation) rather than attempting a broad AI assistant, which keeps scope manageable.