One feature, three agents,no one in anyone’s way.
The lead writes the tasks. You give each one an owner and mark what waits for what. Auto-dispatch sends every ready task to its own agent — Codex on the API, Qwen on the UI — each in an isolated worktree. When the API is done, “Add a reviewer” puts a different model on the same tree.
Backend and frontend in parallel, a blocked task that unblocks itself, a reviewer that doesn’t touch the code.
A lead fills the Task board, you assign owners and a dependency, auto-dispatch hands out the work, and a second model reviews the API.
One task, one owner
Assign a task to a specific agent: it is sent only to that agent, and other agents can’t move or delete it.
Dependencies that hold
A task that depends on another stays blocked until that one is Done — then auto-dispatch sends it out by itself.
A worktree per agent
Parallel agents never edit the same checkout; each branch gets your .env and setup command before its first prompt.
A second model reviews
“Add a reviewer” opens another harness on the same worktree, with a prompt that says: do not edit anything.
How the squad is wired
One board, three agents, one reviewer.
More ways to ship with agents
Every scenario runs on the cards you already saw — pick the next one.
Run your next feature as a squad
Claude Code, Codex and Qwen on one board.
macOS and Linux are coming later.
The first-run wizard installs an agent CLI, Git and the dictation model for you — or skips each step.