chuk-view-schemas

Pydantic v2 schemas and FastMCP decorators for chuk-mcp-ui Views.

Build MCP Apps with interactive UI — charts, maps, tables, dashboards — from Python, with zero JavaScript.

Install

# Schemas only
pip install chuk-view-schemas

# With FastMCP decorator support
pip install chuk-view-schemas[fastmcp]

Quick Start

from mcp.server.fastmcp import FastMCP
from chuk_view_schemas.fastmcp import chart_tool
from chuk_view_schemas.chart import ChartContent, ChartDataset

mcp = FastMCP("my-server")

@chart_tool(mcp, "show_popularity")
async def show_popularity(chart_type: str = "bar") -> ChartContent:
    """Show programming language popularity as an interactive chart."""
    return ChartContent(
        chartType=chart_type,
        title="Language Popularity",
        data=[
            ChartDataset(
                label="Usage %",
                values=[
                    {"label": "Python", "value": 31.0},
                    {"label": "JavaScript", "value": 25.2},
                    {"label": "Rust", "value": 18.1},
                ],
            )
        ],
    )

The @chart_tool decorator handles:

• Registering the ui:// resource URI pointing to the hosted chart view
• Wrapping the return in a CallToolResult with content + structuredContent
• Generating a text fallback for the LLM

Available Decorators

from chuk_view_schemas.fastmcp import (
    map_tool,        # Interactive Leaflet map
    chart_tool,      # Bar, line, pie, scatter, etc.
    datatable_tool,  # Sortable, filterable table
    form_tool,       # JSON Schema driven form
    markdown_tool,   # Rich markdown rendering
    video_tool,      # HTML5 video player
    pdf_tool,        # PDF viewer
    dashboard_tool,  # Multi-panel layout
    split_tool,      # Two-panel side-by-side
    tabs_tool,       # Tabbed panels
    detail_tool,     # Single record display
    counter_tool,    # Big number with delta
    code_tool,       # Syntax highlighted code
    progress_tool,   # Progress bars
    confirm_tool,    # Confirmation dialog
    json_tool,       # JSON tree viewer
    status_tool,     # Traffic light status
    view_tool,       # Generic (any view type)
)

Schemas

All schemas are Pydantic v2 models with camelCase aliases for JSON serialization:

from chuk_view_schemas import (
    ChartContent, ChartDataset,
    MapContent, MapLayer,
    DataTableContent, Column,
    FormContent,
    DashboardContent, Panel,
    SplitContent, SplitPanel,
    TabsContent, Tab,
    MarkdownContent,
    VideoContent,
    PdfContent,
)

View Inference

Automatically suggest the best view for your data:

from chuk_view_schemas import infer_view

suggestion = infer_view([{"lat": 51.5, "lon": -0.1, "name": "London"}])
# ViewSuggestion(view="map", confidence=0.80, reason="Objects contain lat/lon fields")

Hosted Views

All views are pre-built and hosted at versioned URLs on Fly.io:

https://mcp-views.chukai.io/chart/v1
https://mcp-views.chukai.io/map/v1
https://mcp-views.chukai.io/datatable/v1
...

The decorators point at these URLs automatically. Your Python server never serves HTML.

Links

• chuk-mcp-ui repo — 66 pre-built views
• Demo server — 88-tool live demo
• Storybook — Browse all views
• Playground — Try views with your data