---
title: "Task | AgentRQ Glossary"
url: https://agentrq.com/glossary/task
description: "A task in AgentRQ is a discrete unit of work created by an AI agent, with a title, description, status, and conversation thread that lets humans track and collaborate on agent work."
---

[Back to Glossary](https://agentrq.com/glossary/)

# Task

A **task** in AgentRQ is a discrete unit of work created by an AI [agent](https://agentrq.com/glossary/agent). Tasks appear on your [task board](https://agentrq.com/glossary/task-board) and serve as the primary collaboration object between the agent and the human — containing a title, description, current [status](https://agentrq.com/glossary/task-status), and a full conversation thread.

## Anatomy of a Task

| Field | Description |
| --- | --- |
| `title` | Short name for the work item |
| `description` | Detailed description of what the agent is doing |
| `status` | Current state: `pending`, `ongoing`, `completed`, `failed` |
| `chat_id` | Unique identifier for the task's conversation thread |
| `attachments` | Files uploaded by the agent or human |
| `messages` | Full conversation history between agent and human |

## Task Lifecycle

text

```text
pending → ongoing → completed
                  ↘ failed
```

1. **Pending** — Task has been created but the agent hasn't started yet
2. **Ongoing** — Agent is actively working on the task
3. **Completed** — Task is finished successfully
4. **Failed** — Task encountered an unrecoverable error

## Creating Tasks via MCP

[Claude Code](https://agentrq.com/glossary/claude-code) agents create tasks using the `task` [MCP tool](https://agentrq.com/glossary/mcp-tool) with the `create` action:

json

```json
{
  "action": "create",
  "payload": {
    "title": "Refactor auth middleware",
    "body": "Rewriting the JWT validation logic to fix the token expiry bug"
  }
}
```

`help {"action": "task.create"}` lists every field the payload takes. The older `createTask` tool name still works for clients that are already installed.

The task immediately appears on the human's task board with real-time updates.

## Tasks as Collaboration Units

Each task has its own conversation thread. The agent can send messages via `task` reply, and the human can respond directly. This makes tasks the natural unit for [human-in-the-loop](https://agentrq.com/glossary/human-in-the-loop) oversight — one task per work item, with full history preserved.

## Related Terms

- → [Task Board](https://agentrq.com/glossary/task-board)
- → [Task Status](https://agentrq.com/glossary/task-status)
- → [Bidirectional Messaging](https://agentrq.com/glossary/bidirectional-messaging)
- → [Notification](https://agentrq.com/glossary/notification)
- → [Approval](https://agentrq.com/glossary/approval)