AO Fleet Tools

Fleet monitoring, management, and MCP server for AO Agent Orchestrator.

Auto-discovers all AO projects from ~/.ao and shows a live dashboard.

ao-fleet-monitor.sh          # one-shot status
ao-fleet-monitor.sh --watch  # live dashboard (60s refresh)
ao-fleet-monitor.sh --watch 30  # 30s refresh
ao-fleet-monitor.sh --pull   # pull latest for all repos
ao-fleet-monitor.sh --restart # restart crashed daemons
ao-fleet-monitor.sh --full   # pull + monitor

Runs via cron every 5 minutes. Checks all fleet daemons, restarts crashed ones, logs to brain repo.

# Install as cron
(crontab -l; echo "*/5 * * * * /path/to/ao-daemon-watchdog.sh") | crontab -

Runs via cron every 15 minutes. Deletes node_modules, .next, .nuxt, .svelte-kit, .turbo, .vercel, cargo target dirs from ~/.ao worktrees.

# Install as cron
(crontab -l; echo "*/15 * * * * /path/to/ao-worktree-cleanup.sh") | crontab -

mcp/ contains a Node.js MCP server for fleet monitoring and control.

Monitoring:

• fleet.status — health of all daemons
• fleet.agents — active agents across projects
• fleet.queues — queue stats
• fleet.workflows — active/recent workflow runs
• fleet.prs — open/merged PRs
• fleet.tasks — task counts by status

Control:

• fleet.daemon.start — start daemon for a project
• fleet.daemon.stop — stop daemon
• fleet.daemon.restart — restart daemon
• fleet.queue.enqueue — queue a task
• fleet.task.create — create task in any project

cd mcp && npm install
node --experimental-strip-types server.ts

Add to .mcp.json:

{
  "mcpServers": {
    "fleet": {
      "command": "node",
      "args": ["--experimental-strip-types", "path/to/mcp/server.ts"]
    }
  }
}

• AO CLI installed
• gh CLI authenticated
• Node.js 22+

MIT