Create publication-ready plots from an AI assistant

Register the service

Codex users can register the remote Streamable HTTP endpoint with one command:

codex mcp add boxplotr --url https://mcp.chemgrid.org/boxplotr/

Then ask the assistant to call generate_boxplot. Claude Desktop, Antigravity and other remote-MCP clients can use the same URL. No authorization header is required.

Examples generated through the public MCP server

Points and means

Show the observations behind each summary

The bundled five-sample CSV uses custom colours, jittered raw points and red diamonds for sample means. This makes distribution spread and outliers visible alongside medians and quartiles.

show_points: true · add_means: true · custom colors

Example promptUse the BoxPlotR MCP tool with the attached five-sample CSV. Create a vertical ggplot2 box plot titled “Five sample distributions”. Label the axes “Sample” and “Measured value”, use five distinct blue, cyan, teal, amber and red colours, show the individual jittered observations, add sample means, and return a PNG.

Download the tested CSV

Violin plot

Reveal distribution shape

The bundled text scenario is rendered with the ggplot2 engine and Nature-style typography. Violin geometry exposes density while overlaid points retain the individual observations.

plot_type: violin · style_guide: nature · show_points: true

Example promptUse the BoxPlotR MCP tool with the attached three-sample CSV. Create a vertical violin plot with the ggplot2 engine and Nature style. Title it “Distribution shape by sample”, label the y-axis “Measured value”, colour the groups green, orange and purple, overlay the raw observations, omit mean markers, and return a PNG.

Download the tested CSV

Logarithmic scale

Compare measurements spanning orders of magnitude

The supplied Excel example contains Baseline, Treated and Knockout measurements on very different scales. After a checked CSV conversion, logarithmic rendering keeps all three conditions legible.

log_scale: true · style_guide: science · add_means: true

Example promptRead the attached Excel example, use its Baseline, Treated and Knockout columns, and call the BoxPlotR MCP tool. Create a ggplot2 box plot using Science style and a logarithmic y-axis. Title it “Responses across measurement scales”, use blue, red and teal fills, add mean markers without raw points, and return a PNG.

Download the original Excel file · CSV used by MCP

Editorial recreation

Reproduce a public-news chart treatment

This original demonstration follows the visual language of Figure 11a in the public Economist Impact LAC Infrascope 2021/22 report: a restrained background, direct labels and three strongly separated categories.

The values are illustrative, reconstructed from the published chart ranges, and are not the report's underlying dataset. Source: Economist Impact, Figure 11a.

style_guide: economist · custom editorial palette · descriptive axis labels

Example promptUse the BoxPlotR MCP tool with the attached illustrative score data. Recreate the restrained editorial look of Economist Impact Figure 11a without copying the original figure. Make a vertical ggplot2 box plot in Economist style titled “Overall index score by risk-allocation score”. Label both axes clearly, colour Score 0 red, Score 50 grey and Score 100 blue, omit raw points and means, and return a PNG. State that the values are illustrative.

Download the illustrative CSV

Example tool call

{
  "values": "Control,Treatment\n1.2,2.4\n1.5,2.9\n1.8,3.1",
  "plot_type": "boxplot",
  "plot_engine": "ggplot2",
  "style_guide": "nature",
  "title": "Treatment response",
  "x_label": "Condition",
  "y_label": "Response",
  "colors": ["#2563eb", "#ef4444"],
  "show_points": true,
  "add_means": true,
  "output_format": "png"
}

Available modifications include plot type, rendering engine, journal or editorial style, orientation, logarithmic scaling, title and axis labels, group colours, raw-point overlays, mean markers for box plots and output format.

Output attachments and supported options

Set output_format to png, svg or pdf. The public response contains a text summary plus a PNG image or an embedded SVG/PDF resource. Resources include uri, mimeType and a base64 blob; the complete file is embedded in the response. Save the attachment in your client.

The public tool accepts values, plot_type, plot_engine, style_guide, orientation, log_scale, whisker_type, title, x_label, y_label, colors, show_points, add_means, transparent_background, unfilled_shapes and output_format. Use hexadecimal colours such as #2563eb. Styles are none, nature, science, economist and ft.

Transparent figures for Illustrator

Set transparent_background: true to remove page and panel backgrounds, including coloured style backgrounds. Set unfilled_shapes: true to keep shape outlines and statistical lines while removing box, violin and bean area fills. Both options work with PNG, SVG and PDF; use PDF or SVG for editable vector shapes in Illustrator.

{"values":"Control,Treatment\n1,2\n2,3\n3,4\n4,5", "output_format":"pdf", "transparent_background":true, "unfilled_shapes":true}

In the web app, select the matching PDF and SVG download options above the download buttons. In local stdio, put both boolean options under styling.

The separate local stdio server accepts output_path and additional controls for notches, variable widths, mean confidence intervals, subtitles, grids and point styling. These are not parameters of the public endpoint.

Privacy and fair use

Datasets and raw client addresses are not sent to Google Analytics. Client addresses are immediately converted to one-way pseudonymous identifiers for quota enforcement and usage reporting. Temporary generated files are eligible for deletion after one hour and cleaned up on subsequent plot requests.