Multiple AI coding agents need to coordinate work, and I need to observe that easily.
kanban-md is a file-based Kanban board where each task is a Markdown file with YAML frontmatter. There is no server and no database. You can commit the board with your code, review task changes in PRs, and work offline.
What it does
- Manage tasks via CLI (
create,list,show,edit,move,delete,archive,pick,metrics,context) - Interactive TUI (
kanban-md-tui) for human browsing and editing - Atomic
pick --claimworkflow to coordinate parallel agents safely - Classes of service + WIP limits + dependency tracking
- Compact output mode designed for agent context efficiency
Why I made it
In multi-agent coding, race conditions are common: two agents pick the same task, stale claims block progress, and verbose JSON burns tokens. I wanted a Go-native tool that keeps workflows deterministic while staying simple.
Why it is useful vs alternatives
Compared to hosted issue trackers and API-first boards:
- No auth/API dependency: works in local/CI environments without setup friction
- Git-native workflow: tasks are plain files, diff-able and merge-able
- Agent-safe primitives: claim/expiry + atomic pick reduce duplicate work
- Token-efficient output:
--compactis much lighter than full JSON in agent loops - Cross-platform binary distribution: straightforward install for macOS/Linux/Windows
Install
# Homebrew
brew install antopolskiy/tap/kanban-md antopolskiy/tap/kanban-md-tui
# Go install
go install github.com/antopolskiy/kanban-md/cmd/kanban-md@latest
go install github.com/antopolskiy/kanban-md/cmd/kanban-md-tui@latest
Quickstart
# install agent skills
kanban-md skill install
# initialize board
kanban-md init
# create tasks
kanban-md create "Add retry logic" --priority high --tags backend
kanban-md create "Write API tests" --priority medium --tags testing
# atomic claim + move to in-progress
kanban-md pick --claim agent-1 --move in-progress
# token-efficient list for agents
kanban-md list --compact
# optional: visual board for humans
kanban-md-tui