> ## Documentation Index
> Fetch the complete documentation index at: https://docs.toorow.com/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP App & Agent Use Cases Library

> Interactive workflows demonstrating how AI agents trigger tools that render embedded MCP App UI widgets directly inside host client interfaces.

Unlike traditional AI tools that return plain text or raw JSON wall-of-responses, **toorow** features an **MCP App architecture**. When an AI agent executes a toorow FastMCP tool, the host client receives a dual-channel payload that renders **interactive React/TypeScript UI widgets (`ui/widgets/`)** directly inside the chat window.

***

## The MCP App Dual-Channel Architecture

```
+-----------------------------------------------------------------------------------+
|  AI AGENT CHAT INTERFACE (Claude Desktop / Cursor / Antigravity Host)             |
|                                                                                   |
|  User Prompt: "Show me this week's performance and any anomalies across channels"  |
|                                                                                   |
|  +-----------------------------------------------------------------------------+  |
|  |  LLM TEXT CHANNEL (Leans Summary <= 30 lines)                               |  |
|  |  "Overall spend was €12,450 (+8%), generating 1,420 conversions. Meta ROAS   |  |
|  |   improved to 3.8x, while GA4 conversions dipped 12% on July 14..."          |  |
|  +-----------------------------------------------------------------------------+  |
|                                                                                   |
|  +-----------------------------------------------------------------------------+  |
|  |  EMBEDDED MCP APP UI WIDGET (ui/widgets/daily-report)                       |  |
|  |  [ Interactive Performance Chart | Metric Selector | Anomaly Inspector ]    |  |
|  |  - Meta Ads:    €5,200 | 680 conv | ROAS 3.8x [Sparkline: ↗]                |  |
|  |  - GA4 Organic:     0  | 410 conv | CPA  €0.00 [Sparkline: ↘]                |  |
|  |  - TikTok Ads:  €3,100 | 210 conv | ROAS 2.1x [Sparkline: →]                |  |
|  |  * Click any row to inspect underlying pull provenance or context events       |  |
|  +-----------------------------------------------------------------------------+  |
+-----------------------------------------------------------------------------------+
```

***

## Featured MCP App Use Cases

### 1. Executive Morning Briefing App

* **User Intent**: *"Give me my morning marketing briefing."*
* **Agent Action**: Executes `get_morning_briefing(project_id)`.
* **MCP App UI Experience**: Renders an interactive **Morning Briefing Widget**. The executive can toggle date ranges, inspect daily metric deltas, and review flagged anomaly spikes with one click.

***

### 2. Multi-Channel Marketing Attribution & ROI App

* **User Intent**: *"Compare ROI and spend efficiency between Meta Ads, TikTok, and GA4."*
* **Agent Action**: Executes `get_daily_report(report_profile="cross_channel")`.
* **MCP App UI Experience**: Renders a **Cross-Channel Performance Matrix Widget**. Displays side-by-side spend, CPA, and ROAS visualizations computed directly from dbt semantic ratio views (`semantic_roas`, `semantic_cpa`).

***

### 3. Anomaly Root-Cause Inspector App

* **User Intent**: *"Why did GA4 conversions drop on July 14?"*
* **Agent Action**: Executes `get_daily_report` + queries `context_events`.
* **MCP App UI Experience**: Renders an **Anomaly Inspector Widget**. Visually correlates the conversion dip on July 14 with a registered context event (*"Payment Gateway Migration v2"*), allowing the user to drill down into raw pull provenance (`pull_id`).

***

### 4. Interactive Datastream & Mapping Manager App

* **User Intent**: *"Show me all connected datastreams and mapping statuses."*
* **Agent Action**: Executes `flows_list()`.
* **MCP App UI Experience**: Renders an **Admin Datastream Manager Widget**. Enables administrators and agents to inspect datastream health, verify target field mappings (`app.target_fields`), and trigger on-demand backfills.

***

## Next Steps & Cross-References

<CardGroup cols={2}>
  <Card title="MCP Host Integration" icon="network-wired" href="/mcp-host-integration">
    Connect your favorite MCP host client to render toorow MCP App widgets.
  </Card>

  <Card title="Agent Tools Reference" icon="wrench" href="/agent-tools">
    Inspect the FastMCP tools powering MCP App UI payloads.
  </Card>

  <Card title="Agent Rules" icon="scroll" href="/agent-rules">
    Learn how to write LLM rules that preserve `structuredContent`.
  </Card>

  <Card title="Semantic Layer" icon="layer-group" href="/semantic-layer">
    Explore the canonical dbt marts feeding MCP App charts.
  </Card>
</CardGroup>