Add MCP App visualization server for analyzing-data skill

Adds a Prefab-based MCP App server that renders interactive charts and
tables inline in Cursor when the user requests visual output from the
analyzing-data skill.

- New `viz_mcp.py` server with `render_chart` and `render_table` tools
  using Prefab UI (BarChart, LineChart, DataTable with tabs, search,
  sort, pagination)
- Register `analytics-viz` MCP server in both `.claude-plugin/plugin.json`
  and `.cursor-plugin/plugin.json`
- Update SKILL.md with visualization workflow guidance

Author: kaxil

.cursor-plugin/plugin.json

@@ -36,5 +36,15 @@
         }
       ]
     }
+  },
+  "mcpServers": {
+    "analytics-viz": {
+      "command": "uv",
+      "args": [
+        "run",
+        "./skills/analyzing-data/scripts/viz_mcp.py",
+        "--stdio"
+      ]
+    }
   }
 }

skills/analyzing-data/SKILL.md

@@ -42,6 +42,10 @@ Answer business questions by querying the data warehouse. The kernel auto-starts
    ```
 
 6. **Present findings** to user.
+7. **Optional visualization** — If the user asks for chart/table visuals and the `analytics-viz` MCP server is available:
+   - Use `render_table(rows=[...], title="...")` for tabular output
+   - Use `render_chart(rows=[...], x_key="...", y_key="...", chart_type="line|bar", title="...", top_n=<int>, sort_desc=<bool>)` for trend/comparison views
+   - The UI has controls to filter/sort/switch chart type within loaded data. Users can ask in chat for a different row count.
 
 ## Kernel Functions
 
@@ -105,3 +109,19 @@ uv run scripts/cli.py cache clear [--stale-only]  # Clear
 
 - [reference/discovery-warehouse.md](reference/discovery-warehouse.md) — Large table handling, warehouse exploration, INFORMATION_SCHEMA queries
 - [reference/common-patterns.md](reference/common-patterns.md) — SQL templates for trends, comparisons, top-N, distributions, cohorts
+
+## MCP App Visualization (Optional)
+
+When an interactive chart/table is more useful than plain text, use the visualization MCP tools after running SQL with this skill's CLI.
+
+Example workflow:
+
+```bash
+# 1) Run query and print JSON rows
+uv run scripts/cli.py exec "import json; df = run_sql('SELECT ...'); print(json.dumps(df.to_dicts(), default=str))"
+```
+
+Then call MCP tool:
+
+- `render_table(rows=<parsed JSON>, title="Query Results")`
+- `render_chart(rows=<parsed JSON>, x_key="<col>", y_key="<metric>", chart_type="line", title="Descriptive Title")`

skills/analyzing-data/scripts/viz_mcp.py

@@ -0,0 +1,225 @@
+#!/usr/bin/env python3
+# /// script
+# requires-python = ">=3.11"
+# dependencies = [
+#     "fastmcp[apps]>=3.1.0",
+#     "prefab-ui==0.10.0",
+# ]
+# ///
+"""Visualization-only MCP server for the analyzing-data skill.
+
+This server is intentionally small:
+- Query execution and business logic stay in `scripts/cli.py`
+- This server only renders MCP App visualizations (table/chart)
+"""
+
+from __future__ import annotations
+
+import sys
+from collections.abc import Mapping
+from typing import Any, Literal
+
+from fastmcp import FastMCP
+from fastmcp.tools import ToolResult
+from prefab_ui.app import PrefabApp
+from prefab_ui.components import (
+    Column,
+    DataTable,
+    DataTableColumn,
+    Heading,
+    Muted,
+    Tab,
+    Tabs,
+)
+from prefab_ui.components.charts import BarChart, ChartSeries, LineChart
+
+MAX_ROWS = 500
+
+mcp = FastMCP(
+    "Analytics Visualization MCP",
+    instructions=(
+        "Visualization-only server for query result display. "
+        "Use render_table for tabular results and render_chart for line/bar charts. "
+        "Do not run SQL here."
+    ),
+)
+
+
+def _to_json_safe(value: Any) -> Any:
+    """Convert values into JSON-safe types for tool payloads."""
+    if isinstance(value, (str, int, float, bool)) or value is None:
+        return value
+    if isinstance(value, Mapping):
+        return {str(key): _to_json_safe(item) for key, item in value.items()}
+    if isinstance(value, (list, tuple)):
+        return [_to_json_safe(item) for item in value]
+    return str(value)
+
+
+def _normalize_rows(rows: list[dict[str, Any]]) -> tuple[list[dict[str, Any]], bool]:
+    """Normalize rows and enforce a max payload size."""
+    if not isinstance(rows, list):
+        raise ValueError("rows must be a list of objects")
+
+    normalized: list[dict[str, Any]] = []
+    for row in rows[:MAX_ROWS]:
+        if not isinstance(row, Mapping):
+            raise ValueError("each row must be an object/dict")
+        normalized.append(
+            {str(key): _to_json_safe(value) for key, value in row.items()}
+        )
+
+    return normalized, len(rows) > MAX_ROWS
+
+
+def _build_table_app(title: str, rows: list[dict[str, Any]]) -> PrefabApp:
+    """Build a Prefab DataTable app."""
+    keys = list(dict.fromkeys(k for row in rows for k in row))
+    columns = [DataTableColumn(key=k, header=k, sortable=True) for k in keys]
+
+    with Column(gap=4, css_class="p-6") as view:
+        Heading(title)
+        Muted(f"{len(rows)} rows")
+        DataTable(
+            columns=columns,
+            rows=rows,
+            searchable=True,
+            paginated=True,
+            page_size=20,
+        )
+
+    return PrefabApp(view=view)
+
+
+def _build_chart_app(
+    title: str,
+    rows: list[dict[str, Any]],
+    x_key: str,
+    y_key: str,
+    chart_type: str,
+    top_n: int | None,
+    sort_desc: bool,
+) -> PrefabApp:
+    """Build a Prefab chart app with tabs for chart and table views."""
+    if sort_desc:
+        rows = sorted(rows, key=lambda r: _numeric(r.get(y_key)), reverse=True)
+    else:
+        rows = sorted(rows, key=lambda r: _numeric(r.get(y_key)))
+
+    display_rows = rows[:top_n] if top_n and top_n > 0 else rows
+
+    Chart = BarChart if chart_type == "bar" else LineChart
+    keys = list(dict.fromkeys(k for row in rows for k in row))
+    columns = [DataTableColumn(key=k, header=k, sortable=True) for k in keys]
+
+    with Column(gap=4, css_class="p-6") as view:
+        Heading(title)
+        Muted(
+            f"Showing {len(display_rows)} of {len(rows)} rows · Ask in chat for a different count"
+        )
+        with Tabs():
+            with Tab("Chart"):
+                Chart(
+                    data=display_rows,
+                    series=[ChartSeries(data_key=y_key, label=y_key)],
+                    x_axis=x_key,
+                    show_legend=True,
+                )
+            with Tab("Table"):
+                DataTable(
+                    columns=columns,
+                    rows=display_rows,
+                    searchable=True,
+                    paginated=True,
+                    page_size=20,
+                )
+
+    return PrefabApp(view=view)
+
+
+def _numeric(value: Any) -> float:
+    """Extract a numeric value for sorting, defaulting to -inf."""
+    if isinstance(value, (int, float)):
+        return float(value)
+    if isinstance(value, str):
+        try:
+            return float(value.replace(",", "").replace("$", ""))
+        except ValueError:
+            pass
+    return float("-inf")
+
+
+@mcp.tool(app=True)
+def render_table(
+    rows: list[dict[str, Any]],
+    title: str = "Query Results",
+) -> ToolResult:
+    """Render rows in an interactive table MCP App.
+
+    Args:
+        rows: Array of row objects from a query
+        title: Table title
+    """
+    normalized, truncated = _normalize_rows(rows)
+    summary = f"{title}: {len(normalized)} rows"
+    if truncated:
+        summary += f" (truncated from {len(rows)})"
+    return ToolResult(
+        content=summary,
+        structured_content=_build_table_app(title, normalized),
+    )
+
+
+@mcp.tool(app=True)
+def render_chart(
+    rows: list[dict[str, Any]],
+    x_key: str,
+    y_key: str,
+    chart_type: Literal["bar", "line"] = "line",
+    title: str = "Query Chart",
+    top_n: int | None = None,
+    sort_desc: bool = True,
+) -> ToolResult:
+    """Render rows as a line/bar chart in an MCP App.
+
+    Args:
+        rows: Array of row objects from a query
+        x_key: Column to use as labels (x-axis)
+        y_key: Column to use as metric values (y-axis)
+        chart_type: "line" or "bar"
+        title: Chart title
+        top_n: Show only top N rows (default: show all)
+        sort_desc: Sort by metric descending (default: True)
+    """
+    normalized, truncated = _normalize_rows(rows)
+    available_keys = sorted({key for row in normalized for key in row})
+
+    if x_key not in available_keys:
+        raise ValueError(f"x_key '{x_key}' not found. Available keys: {available_keys}")
+    if y_key not in available_keys:
+        raise ValueError(f"y_key '{y_key}' not found. Available keys: {available_keys}")
+
+    display_count = min(top_n, len(normalized)) if top_n else len(normalized)
+    summary = f"{title}: {display_count} of {len(normalized)} rows, {chart_type} chart by {y_key}"
+    if truncated:
+        summary += f" (truncated from {len(rows)})"
+
+    return ToolResult(
+        content=summary,
+        structured_content=_build_chart_app(
+            title=title,
+            rows=normalized,
+            x_key=x_key,
+            y_key=y_key,
+            chart_type=chart_type,
+            top_n=top_n,
+            sort_desc=sort_desc,
+        ),
+    )
+
+
+if __name__ == "__main__":
+    if "--stdio" in sys.argv:
+        mcp.run(transport="stdio")
+    else:
+        mcp.run(transport="stdio")