Complete Task
complete_task is one of the 32 MCP tools in Weft, the AI task manager you run by talking to Claude, Cursor, Codex or ChatGPT through one MCP endpoint. The AI calls this when the work is actually finished — the card moves to Done with completion time recorded, and the optional outcome (what was done, where the result lives) is appended to the card as an "Outcome (AI, date)" note. No column id needed. Work marked as needing you does NOT move to Done: it goes to your queue with the receipt attached, and if the agent said what it would do next, accepting approves exactly that step and hands it back to the agent. If the task carries a verification plan, the server executes its checks right here — fetching the URL, looking for the quote, comparing the hash: an executable/state_delta plan whose blocking checks all pass closes verified by the server; a failed blocking check keeps the card out of Done and shows the agent what the server observed.
Weft es el gestor de tareas con IA para fundadores en solitario y equipos pequeños.
La referencia de herramientas se mantiene en inglés: los nombres de las herramientas, los argumentos y los ejemplos son idénticos en todos los idiomas.
Argumentos
| Nombre | Tipo | Obligatorio | Descripción |
|---|---|---|---|
| id | uuid | sí | Task id to complete. |
| outcome | string | no | 1-3 sentences: what was done and where the result lives. |
| artifacts | { kind, uri }[] | no | Evidence a person could check without trusting the account of it — a URL, a hash, a message id. |
| learnings | (string | { trigger, content })[] | no | Up to 3 lessons worth remembering beyond this task — they become memory notes and surface in every future get_my_work, stamped with source and date. |
| nextStep | string | no | What the agent would do next if you accept (≤200 chars) — shown above the Accept button, so accepting approves exactly that. |
Devuelve
The task, now in Done with completedAt set — or awaitingReview:true when it stopped at you instead.
Anotaciones
Herramienta de escritura por defecto. La herramienta modifica el estado; los clientes MCP pueden llamarla libremente en cuanto sepan qué quiere el usuario. Que el cliente pida confirmación o no es decisión del cliente; la especificación del Model Context Protocol recomienda mantener a una persona en el circuito, con la posibilidad de denegar invocaciones de herramientas.
Ejemplo
Lo que dice el usuario:
"Mark the deploy task done — released in v1.2.0."Relacionado
- Las 32 herramientas MCP — la vista de catálogo.
- Shuttle — cómo el cliente de IA usa estas herramientas.
- Llama a esta herramienta desde tu propio cliente: consulta las guías de configuración o accede directamente a
https://letsweft.com/api/mcp.