software · Picked by us
A board that keeps coding agents in reviewable states
A local kanban board and CLI for agent-led coding: briefs, isolated worktrees, recorded evidence and explicit review gates stay in plain Markdown — and the merge stays with the human.
By 
Vladimir Karataev @shinKatana0
The maker explains the workflow and how the agents are governed
npx briefboard init, then npx briefboard serve — MIT and local
What it does
Run long coding-agent jobs without losing decisions, ownership or state in chat — every task moves through brief, implementation and review with evidence attached.
Who it helps
Developers running multiple coding agents who want state, review and rollback outside the chat window.
How it was made
- Models
- Claude Code · creator-reported
- Tools
- Claude Code · Primary build agentNode.js · Runtime for the CLI and local board
- Work is planned as briefs and moved through backlog, brief, implementation, review and decision states; Markdown stays the source of truth.
- Each task can run in an isolated worktree with declared dependencies, and review records the evidence before anything merges.
- The maker reports Claude Code was the primary agent used to build the workflow — and to exercise it on real projects.
What the person did
Vladimir Karataev owns the planning, briefs, review, measurements and merges while coding agents write much of the code and prose.
What we know
Creator-reported: A public repository with the full workflow, a maker-authored build note, and a creator profile that links the project.
Access checked — 2026-09-27 · Opened the repository, the maker build note and the creator profile on 2026-09-27; the scout dossier verified the same links the previous day.
Can you use it?
The project is MIT-licensed and stays on GitHub; we link rather than host captures of it. The cover is an original Made With Models illustration, not a screenshot of the tool.
About this project
Picked by Made With Models. Added 2026-09-27. Updated 2026-09-27.