Agent Mission Control -- Supervise the Swarm

Launching more agents is easy. Knowing which one is blocked, what it changed, and whether its work is safe to integrate is the real job.

Agent Mission Control turns parallel coding into an observable operation. It combines worktree isolation, explicit file ownership, pane-state triage, evidence receipts, conflict handling, and bounded recovery into one supervisory workflow.

What you get

• A mission-board format for every active lane
• Non-overlapping ownership and dependency planning
• Working, idle, blocked, awaiting-approval, and review-ready state handling
• Required changed-file, command, result, limitation, and commit receipts
• Integration verdicts based on combined checks rather than agent confidence
• Recovery that preserves logs and avoids broad process killing

Use it when one coding agent is no longer enough but a swarm would otherwise become harder to understand than the project itself.

Pairs with Evidence Gate for claim verification and Gauntlet for enforcement at write, commit, and CI boundaries. It does not silently spend provider money, merge every branch, or bypass human approval at destructive and production edges.