Charted

What marzukia/charted MCP server does

The marzukia/charted MCP server gives AI clients a local interface for generating charts from structured data. It supports 15 chart types: bar, column, line, scatter, pie, area, radar, box plot, histogram, heatmap, Gantt, bubble, combo, polar area, and Sankey. The underlying charted library can produce SVG natively and supports PNG through an optional extra. Other documented output forms include HTML and data URLs.

The server is useful for agents that need to visualize measurements, comparisons, distributions, schedules, flows, or categorical values. It can work with JSON data through chart creation and includes a dedicated CSV charting tool. Built-in themes and custom theme composition let callers control the appearance of generated charts.

How it works

The server starts as a local stdio process. An MCP-compatible client launches charted-mcp, and the client communicates with it through the process rather than through a hosted endpoint. The package can be run in an isolated temporary environment with uvx, so charted does not need to be added to an existing Python project or virtual environment for MCP use.

Four tools are documented: create_chart, list_chart_types, list_themes, and chart_from_csv. The discovery tools can be used to inspect supported chart types and available themes before creating a visualization. Charted itself has no runtime dependencies; PNG generation and MCP support are optional features with their own dependencies.

Setup and configuration

Run the marzukia/charted MCP server with:

uvx --from charted[mcp] charted-mcp

For a client configuration, use the equivalent stdio entry:

{
  "mcpServers": {
    "charted": {
      "command": "uvx",
      "args": ["--from", "charted[mcp]", "charted-mcp"]
    }
  }
}

The README specifically documents setup for Claude Code, Cursor, and Cline. Claude Code can also register the server with its claude mcp add command. No API key, account, or remote service is specified for the MCP setup.

If you are using charted as a Python library instead of MCP, install it with uv add charted or pip install charted. PNG export requires the optional PNG dependency, while the base library remains pure Python.

Tools and capabilities

• create_chart creates a chart from supplied data and chart settings.
• chart_from_csv creates a chart from CSV input.
• list_chart_types reports the chart types available to the client.
• list_themes reports the built-in themes.
• Multiple series are supported, including stacked, grouped, and side-by-side layouts.
• Negative values receive zero-baseline handling.
• Charts can be styled with themes and per-series style builders.
• The library also supports markdown export, a command-line interface, and inline rendering in Jupyter.

Limitations and notes

The provided material does not document authentication, remote hosting, environment variables, or a specific software license. PNG output is not part of the dependency-free base path; it requires the optional cairosvg dependency. The MCP command uses the mcp package extra to install server support. Client compatibility beyond the named Claude Code, Cursor, and Cline configurations is not established by the supplied documentation.