local-first mission control for your AI agent workforce Β· localhost only
π Today
Loadingβ¦
π Pending my approval (Inbox β cockpit/TG first answer wins, auto-deny on timeout)
RUNS
β GATE β item-by-item approval (π΄ = high risk)
approve = release and continue / reject = block (every item needs an explicit decision before submitting)
π GATE β this flow is waiting for your approval
(select a run or press βΆ Run)
Loadingβ¦
Open issues (stored in data/issues.json; add via python -c "import issues_store; issues_store.add('title')")
Loadingβ¦
β unsavedDrag elements from the left onto the canvas β connect them β double-click a node to edit params β Save β Run (becomes a Run)
Elements (click to add)
Node
Employees (each line = a bot with its own independent memory)
Loadingβ¦
β Pick an employee on the left to start chatting
After a reboot, all your project CLI windows are listed here β hit π Reopen to bring one back (full context intact). π Pin the lines you run often so they show first after reboot.
π Pinned lines (open these first after reboot)
β
All recent project sessions
Loadingβ¦
β¨ οΌNew task
Predefined actions (after launching, switch back to Ops to watch the log)
Loadingβ¦
Free-form command β instruct the AI (claude headless, prompt via stdin)
Uses your Max subscription quota (server blocks ANTHROPIC_API_KEY); check "Allow tools" = SquadCue may modify files, unchecked = read-only analysis