How Grapthway Flow works
plan with AI · ship with agents
1 · Create a project & board
Start a kanban self-project (not an AI-Agent VPS project) and a board. Every board ships with the canonical Active-Sprint columns — READY FOR DEV → IN DEV → READY TO TEST → TESTING → READY FOR SHOW → READY FOR BUSINESS SIGN OFF, plus ON-HOLD — and a Backlog.
2 · Fill the backlog
Add stories, tasks, bugs, documentation, or custom work-items. Each gets an issue key (PROJ-123). An item can have sub-tasks — a parent can’t move right until its sub-tasks reach READY FOR SHOW, and moving a parent to sign-off cascades its sub-tasks there too.
3 · The AI does the work
Drop an item into READY FOR DEV (or hit “Run with AI”). An AI-Chat manager room plans it and spawns a sub AI-Chat executor; a worker walks the item forward — building in IN DEV, validating in TESTING — until it reaches READY FOR SHOW for your review. One manager room per project keeps everything in the same place.
4 · You stay in control
Every column has a gate: auto (the worker advances it) or user (it parks for you). The usual setup is user-drops-to-DEV, auto through TESTING, then a human gate at READY FOR SHOW — or gate every step. ON-HOLD parks anything. Nothing ships to production without your sign-off.
5 · Repos & subscription
Bind an item to a repository + git credential (the same ms-container repos the AI-Agent uses) or let the worker discover it. Your Flow plan sets how many agents can run at once and how many projects you can have — configurable by an admin; agent pickup pauses when a subscription lapses.