Overview

These docs follow unreleased main. The latest published TanStack Charts release is 0.7.0; it is pre-alpha and its API may change between releases.

TanStack Charts is a small, framework-agnostic chart grammar for TypeScript and JavaScript. Give each mark its natural data, map fields or accessors to visual channels, and use compact scales to define common numeric and categorical axes. TanStack Charts compiles that declaration into a responsive, keyed scene and renders accessible SVG by default, with Canvas available as an opt-in surface.

TanStack Charts builds on the grammar-of-graphics tradition established by Leland Wilkinson and developed through projects such as ggplot2, Vega-Lite, and Observable Plot. Observable Plot is the closest API influence for mark-local data, channels, and layered composition. TanStack Charts applies those ideas to typed application infrastructure with explicit scale and algorithm boundaries, responsive scene compilation, and framework lifecycle.

The library is designed for two equally important authors:

• People should get polished, responsive charts from a short declaration.
• AI should be able to compose, inspect, and modify charts without learning an application-specific series model or guessing at hidden behavior.

The same definition can feed the vanilla DOM host and framework adapters. React and Octane also provide optional Canvas entries; the experimental React Native adapter consumes definitions from the universal entry.

A chart is a composition#

import { defineChart, differenceY, window } from '@tanstack/charts'
import { scaleLinear } from '@tanstack/charts-scales/linear'
import { scaleUtc } from 'd3-scale'

interface ClosingPrice {
  Date: Date
  Close: number
}

const observations: readonly ClosingPrice[] = [
  { Date: new Date('2013-05-13T00:00:00Z'), Close: 64.96 },
  { Date: new Date('2013-05-14T00:00:00Z'), Close: 63.41 },
  { Date: new Date('2013-05-15T00:00:00Z'), Close: 61.26 },
  { Date: new Date('2013-05-16T00:00:00Z'), Close: 62.08 },
  { Date: new Date('2013-05-17T00:00:00Z'), Close: 61.89 },
  { Date: new Date('2013-05-20T00:00:00Z'), Close: 63.28 },
  { Date: new Date('2013-05-21T00:00:00Z'), Close: 62.81 },
  { Date: new Date('2013-05-22T00:00:00Z'), Close: 63.05 },
]

const rows = window(observations, {
  size: 3,
  orderBy: 'Date',
  anchor: 'end',
  partial: false,
  outputs: {
    average: { value: 'Close', reduce: 'mean' },
  },
})

const closingPriceChart = defineChart({
  marks: [
    differenceY(rows, {
      x: 'Date',
      y1: 'average',
      y2: 'Close',
      positiveFill: '#16a34a',
      negativeFill: '#dc2626',
      stroke: '#166534',
      comparisonStroke: '#475569',
    }),
  ],
  x: {
    scale: scaleUtc,
    nice: true,
    axis: { label: 'Date' },
  },
  y: {
    scale: scaleLinear,
    nice: true,
    grid: true,
    axis: { label: 'Close (USD)' },
  },
})

The rolling average is an ordinary, visible window transform. differenceY owns sign segmentation and exact crossing interpolation. The y axis uses the compact numeric scale; the calendar-aware x axis upgrades only that mapping to D3's scaleUtc. Installation lists the exact packages, and Scales explains the ownership boundary.

What TanStack Charts owns#

TanStack Charts owns the parts that make a declarative chart reliable inside an application:

• A grammar of typed marks, channels, scales, guides, and layers
• Container-responsive pixel ranges and automatic guide margins
• A renderer-neutral scene with stable keys
• Default SVG rendering, keyed DOM reconciliation, optional Canvas painting, and interruptible animation
• Pointer and keyboard focus, selection callbacks, and native tooltips
• Framework-agnostic runtime state with thin framework adapters
• Light and dark mode defaults based on inherited color and CSS variables
• Public extension points for custom marks, focus strategies, spatial indexes, and renderers

Keep ownership explicit#

TanStack Charts keeps data preparation explicit and optional algorithms behind narrow imports.

Prepared data can come from TanStack transforms, D3, SQL, a server, or ordinary TypeScript; marks consume it without requiring a special series container.

Defaults for the common case#

The normal path is intentionally short:

• Omit width to follow the chart container.
• Omit margin to measure axes, tick labels, rotation, and titles automatically.
• Supply ariaLabel; keyboard focus is enabled by default.
• Add the tooltip extension when a native value tooltip is enough.
• Start with compact linear, band, point, and ordinal scales; upgrade only the scale that needs fuller D3 semantics.
• Let built-in marks infer stable identity from IDs or unique positions; supply a stable key when that identity is unavailable or can change.
• Let field names, datum types, scales, interaction points, and adapters infer without casts.
• Use inherited currentColor and the --ts-chart-* CSS variables for automatic theme integration.

Every automatic behavior has an explicit escape hatch. The Guides cover those controls by task rather than repeating the API reference.

Packages#

All packages are ESM and tree-shakeable. Built-in marks and optional capabilities also have subpath exports when a library or design system needs tighter bundle boundaries.

Where to go next#

• Compare Libraries — evaluate Chart.js, Apache ECharts, Recharts, Observable Plot, and TanStack Charts against the pinned evidence.
• Installation — install the core, compact scales, an adapter, and any advanced D3 modules your charts import.
• Quick Start — define, mount, update, and destroy a responsive chart.
• Grammar of Graphics — understand how data, marks, channels, scales, and layers fit together.
• Choosing a Chart — start from the analytical question.
• Example Gallery — browse complete, embeddable compositions.
• Migrating — preserve semantics and establish parity before removing an existing renderer.
• AI Authoring — give an agent the smallest reliable path from intent to verified output.