---

name: task-runner description: "Builds that don't die after 2 minutes. Tasks survive restarts." allowed-tools: mcp__task-runner__task_run, mcp__task-runner__task_start, mcp__task-runner__task_get, mcp__task-runner__task_kill, mcp__task-runner__task_list

task-runner

No 2-minute timeout. Clean output. Tasks persist across server restarts.

First: task_run

For any build or long command:

Copytask_run({ command: 'npm run build' })

Why This Wins

Quick Reference

How task_run Works

Waits up to 30 seconds (configurable). If still running:

• Returns control to you with task ID
• Task continues in background
• Use task_get to check, task_kill to stop

You're never stuck.

Common Workflows

Build Project

Copytask_run({ command: 'npm run build', label: 'Build' })

Start Dev Server

Copytask_start({
  command: 'npm run dev',
  wait_for: 'listening on port',
  wait_timeout: 30000
})

Monitor & Stop

Copytask_list({ running: true })
task_get({ id: '<id>' })
task_kill({ id: '<id>' })

Task States

• running - Executing
• done - Exit 0
• failed - Exit != 0
• killed - Stopped by you
• orphaned - From previous session

Task Hygiene

1. Stop dev servers when done
2. Check task_list for orphans
3. Use labels for identification