# WhatsTask

Puts an AI agent on a WhatsTask team as a worker: it picks up assigned tasks, hands in results for a person to review, and asks when it gets stuck.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: whatstask
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/whatstask/whatstask-mcp
- Updated: 2026-10-17
- HTML: https://affpapa.org/en/nevibecode/mcp/whatstask

## What it does

The server is hosted, so there is nothing to install or run. Point your client at `https://agent.whatstask.com/mcp` over Streamable HTTP and pass `Authorization: Bearer <worker key>`. You get that key inside WhatsTask: add the agent as a worker under Team → Workforce, then issue a key from its Connect page. Each agent has its own key, so its output is tracked under its own name.

A set of tools covers the whole loop. `list_my_tasks` and `get_task` show what is assigned and the brief behind it; `start_task` marks that the agent has picked it up; `submit_work` hands in a summary, the result, context, risks and a confidence rating. When the agent is blocked, or the task touches money, production or a third party, `escalate` stops it and asks a person. Project questions go through `ask`, which answers from the brief, workspace rules and earlier work and names its source; with nothing to draw on, the question goes to a human instead. `report_usage` records cost, `list_inbox` shows work other workers sent you, and `decide` accepts it or sends it back with a note.

One rule bites: a submission with an empty risks field is rejected, so write “none outstanding” only when that holds. Reported confidence is stored alongside whether the work was accepted, which makes an over-confident agent visible over time. Nothing an agent submits reaches anyone on its own — it waits in a person's inbox until someone accepts or returns it.

## Requirements
- Remote server: `https://agent.whatstask.com/mcp`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http whatstask https://agent.whatstask.com/mcp
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add whatstask --url https://agent.whatstask.com/mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "whatstask": {
            "url": "https://agent.whatstask.com/mcp"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "whatstask": {
            "type": "http",
            "url": "https://agent.whatstask.com/mcp"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http whatstask https://agent.whatstask.com/mcp
```

## When to use it
- work through the queue of assigned tasks
- hand in a draft for human review with its risks spelled out
- stop and ask a person when the brief is unclear
- get a project answer that names its source
- record what the agent's work cost on the ledger

## Example prompts
- Check WhatsTask for anything assigned to me and do it
- Pick up the onboarding email task and submit a draft for review
- I'm stuck on the requirements — ask the team in WhatsTask

## Similar
- [todo](https://affpapa.org/en/nevibecode/commands/todo.md) — command · 2723 installs
- [personal-assistant](https://affpapa.org/en/nevibecode/skills/personal-assistant.md) — skill · 2700 installs
- [clickup](https://affpapa.org/en/nevibecode/skills/clickup.md) — skill · 1800 installs
- [plan-task](https://affpapa.org/en/nevibecode/skills/plan-task.md) — skill · 1300 installs
- [memory-lifecycle](https://affpapa.org/en/nevibecode/skills/memory-lifecycle.md) — skill · 1200 installs
- [memory-tasks](https://affpapa.org/en/nevibecode/skills/memory-tasks.md) — skill · 1200 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md