Qué hace
Stop work on a task and keep every agent — including you, on the next run — off it until a human resumes it. Use it when you are STUCK in a way that no question would fix: you are going round the same loop, a credential or access you need isn't there, the task as written can't be done, or you can see yourself about to spend a lot for nothing. Pausing is the honest end of a run that isn't going anywhere, and it is the only tool here that stops the work from being picked straight back up: comment and the next agent claims the same task and hits the same wall, decline_task and it goes back in the queue for someone else to hit. Prefer ask_decision when you have an actual QUESTION — a fork a human could settle in one line — because that keeps the task alive and answerable; pause when you don't even have a question to ask. Prefer decline_task when the task is fine and simply isn't yours. reason is required and is the whole value of the call: it is what the human reads when they come back to it, so say what you hit and what you had already tried, not just "stuck". You cannot un-pause: resuming is a human action in the app, deliberately, so that an agent which paused itself for looping can't put itself back in the loop.
Parámetros
| Nombre | Tipo | Obligatorio | Descripción |
|---|---|---|---|
id | string | Obligatorio | |
reason | string | Obligatorio | What you hit, and what you'd already tried. Posted to the task's thread and shown on the board. |
Anotaciones
Las pistas que el servidor envía con esta herramienta. Un cliente las lee para decidir si te pregunta antes de llamarla.
- Solo lectura
readOnlyHint: false - Cambia datos de tu cuenta de Done.
- Destructiva
destructiveHint: false - Solo agrega o lee; no quita ni sobrescribe nada de lo que ya existe.
- Idempotente
idempotentHint: true - Llamarla dos veces con los mismos argumentos deja el mismo resultado que llamarla una vez.
- Mundo abierto
openWorldHint: false - Solo toca tu propia cuenta de Done.
Qué tokens pueden llamarla
Una conexión hecha con OAuth tiene acceso completo. Un token de Ajustes → Acceso de agentes → Tokens de MCP adicionales se puede acotar, y un token acotado nunca ve más allá de su proyecto o espacio de trabajo.
- Acceso completoPermitido
- Token de solo lecturaRechazado
- Token de un proyectoRechazado
- Token de un espacio de trabajoPermitido
Pruébala
Con Done conectado, pídele a tu asistente algo como:
“Si te sale el mismo error dos veces en esa tarea, pausa la tarea y dime por qué.”
Conecta Done a tu asistente
Un endpoint MCP por cuenta, con OAuth o con un token. Configuración paso a paso para cada cliente: