Create data visualizations by describing what you want. No syntax to memorize.

ggterm is a Grammar of Graphics toolkit for Claude Code. Describe your data and what you want to see in natural language — ggterm handles scales, legends, colors, and layout automatically. Plots appear instantly in a live browser panel alongside your terminal.

mkdir my-analysis && cd my-analysis
npx ggterm-plot setup

That's it. This installs everything, generates a welcome plot, opens the live viewer in your browser, and starts the server. Now talk to Claude:

You: Load the iris dataset and show me sepal length vs petal length, colored by species

Claude: [Creates scatter plot — appears instantly in the live viewer]

You: Add a trend line and style it like Nature journal

Claude: [Updates the plot in place with regression line and Nature styling]

You: Export as PNG for my paper

Claude: [Generates publication-ready output]

ggterm includes specialized visualizations across domains:

See the full Geometry Reference.

Start exploring immediately with bundled datasets — no files needed:

You: Plot the iris dataset, sepal length vs petal length, colored by species

You: Create a volcano plot of the airway DESeq2 results

You: Show Kaplan-Meier survival curves for the lung data, stratified by sex

airway: DESeq2 differential expression results from Himes et al. 2014 (human airway smooth muscle cells). lung: NCCTG lung cancer survival data (Loprinzi et al. 1994).

Or bring your own CSV, JSON, or JSONL files.

ggterm runs as a companion to Claude Code. The live viewer sits in a browser panel (or Wave terminal side panel) and automatically displays every plot you create:

• Instant display — plots appear as interactive Vega-Lite visualizations (tooltips, zoom, pan)
• Command palette — press Cmd+K to fuzzy-search 80+ commands, all 66 geom types, style presets, and export actions
• Plot history — every plot saved automatically; press h for the history sidebar, arrow keys to browse, Home/End to jump
• Help panel — press ? for a 5-tab reference (Getting Started, Plot Types, Shortcuts, Styles, Export)
• Data inventory — on setup, ggterm scans your directory for CSV, TSV, JSON, and JSONL files, analyzes column types and ranges, and writes a catalog so Claude already knows your data before you ask
• Style in place — apply publication presets (Wilke, Tufte, Nature, Economist, APA) without re-running; viewer updates live
• Export — press s/p for SVG/PNG directly from the viewer, or generate standalone HTML

Under the hood, ggterm is a comprehensive Grammar of Graphics implementation — the same foundation as R's ggplot2. Every plot is represented as a declarative PlotSpec that can be rendered to terminal ASCII art or converted to Vega-Lite for the browser. 66 geometry types, 75 scales, statistical transforms, faceting, and themes — all composable through the grammar.

See complete AI-driven workflows with real data:

mkdir my-analysis && cd my-analysis
npx ggterm-plot setup    # Init + welcome plot + open browser + start server

npm install @ggterm/core
npx ggterm-plot init     # Install Claude Code skills
npx ggterm-plot serve    # Start live viewer

npx ggterm

ggterm> .data iris
Loaded Iris dataset: 150 rows, 5 columns

ggterm> gg(data).aes({x: "sepal_length", y: "petal_length", color: "species"}).geom(geom_point())

              Sepal vs Petal Length
  7.0 ┤                         ■■■■
      │                      ■■■■■■■■
  6.0 ┤                   ■■■■■■■■
      │               ●●●●●■■■■■
  5.0 ┤            ●●●●●●●●●
      │         ●●●●●●●●
  4.0 ┤       ●●●●●●
      │     ●●●●
  3.0 ┤   ●●●
      │
  2.0 ┤▲▲▲▲▲▲
      │▲▲▲▲▲▲▲
  1.0 ┤▲▲▲▲
      └──────────────────────────────────
       4.5   5.0   5.5   6.0   6.5   7.0

      ▲ setosa  ● versicolor  ■ virginica

Use the programmatic API directly:

import { gg, geom_point, scale_color_viridis } from '@ggterm/core'

const plot = gg(data)
  .aes({ x: 'sepal_length', y: 'petal_length', color: 'species' })
  .geom(geom_point())
  .scale(scale_color_viridis())
  .labs({ title: 'Iris Dataset' })

console.log(plot.render({ width: 80, height: 24 }))

Every plot is a backend-agnostic PlotSpec — a JSON-serializable specification consumed by the terminal renderer or Vega-Lite exporter. New backends can be added without modifying the grammar layer. See the Architecture for details.

• Quick Start Guide
• Geometry Reference — all 66 plot types
• Live Viewer Guide — command palette, history, keyboard shortcuts
• Architecture — PlotSpec, backends, layer model
• API Reference
• Migration from ggplot2
• Contributing

MIT