Zero-dependency charts for Python and AI agents. SVG and PNG, 15 chart types, a one-line MCP server, and a Claude skill.

Quickstart

uv add charted

Or with pip:

pip install charted

from charted import BarChart

chart = BarChart(
    title="Sales by Quarter",
    data=[120, 180, 210, 150],
    labels=["Q1", "Q2", "Q3", "Q4"],
)
chart.save("chart.svg")
chart.save("chart.png")  # PNG export (requires cairosvg)

That is the whole loop: pass a list of numbers, get back an image. No numpy, no
pandas, no config files.

Core principle: charted itself has zero runtime dependencies. PNG export
and MCP server support are opt-in extras that pull in their own dependencies,
and the base library stays pure Python.

Gallery

Every chart type, rendered. Each image below was generated by charted itself
from the example script; the Quick Tour
shows the code behind each one.


Bar	Column	Line
Scatter	Pie	Area
Radar	Box Plot	Histogram
Heatmap	Gantt	Bubble
Combo	Polar Area

Sankey diagrams are also supported; see the Quick Tour for the
flow and funnel examples.

Integration

Run the MCP server with no install using uvx:

uvx --from charted[mcp] charted-mcp

This fetches charted with the MCP extra into a throwaway environment and starts
the server over stdio, so an agent can generate charts without you adding charted
to a project or virtualenv.

Client config

Client	Setup
Claude Code	`claude mcp add charted -- uvx --from charted[mcp] charted-mcp`
Cursor, Cline, other clients	Add the JSON below to the client's MCP server config

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

The server exposes create_chart, list_chart_types, list_themes, and
chart_from_csv. See the MCP Server section
below for tool details and the pip install path.

Claude Skill

Install the chart Skill in one line:

claude skill install charted

Chart gallery

The Quick Tour renders every chart type with the code that
produced it. The same examples live in docs/examples/.

Why Charted?

Zero runtime dependencies: pure Python, no numpy/pandas required
15 chart types: Bar, Column, Line, Scatter, Pie, Area, Radar, Box Plot, Histogram, Heatmap, Gantt, Bubble, Combo, Polar Area, Sankey
Multi-series support: stacked, side-by-side, grouped layouts
Negative values handled: proper zero baseline calculations
SVG and PNG output: SVG natively, PNG via optional cairosvg (pip install charted[png])
Theme system: 3 built-in presets + custom theme composition
Per-series styling: granular control with SeriesStyle builders
Data loading: CSV/JSON parsers built-in
Markdown export: generate embed-ready markdown snippets
CLI included: create charts without writing Python code
Jupyter ready: charts render inline automatically
Base Chart class: unified API for dynamic chart type selection

Quick Tour

Every chart type shares the same simple interface: pass data, labels, dimensions, and a title:

from charted.charts import BarChart, LineChart, PieChart

# Bar: single series with negatives
BarChart(
    title="Profit/Loss by Region ($M)",
    data=[-12, 34, -8, 52, -5, 28, 41, -19, 15, 60],
    labels=["North", "South", "East", "West", "Central", "Pacific", "Atlantic", "Mountain", "Plains", "Metro"],
    width=700, height=500,
).save("bar.svg")

# Bar: multi-series side-by-side
BarChart(
    title="Revenue vs Expenses by Quarter ($K)",
    data=[[120, -45, 180, -30, 210, -60], [-80, -20, -95, -15, -110, -25]],
    labels=["Q1 Prod", "Q1 Ops", "Q2 Prod", "Q2 Ops", "Q3 Prod", "Q3 Ops"],
    width=700, height=500,
).save("bar_multi.svg")

# Bar: stacked
BarChart(
    title="Budget by Department ($K)",
    data=[[100, -50, 120], [80, 60, -40]],
    labels=["Q1", "Q2", "Q3"],
    series_names=["Revenue", "Expenses"],
    x_stacked=True, width=700, height=400,
).save("bar_stacked.svg")

# Bar: side-by-side with negatives
BarChart(
    title="Revenue vs Expenses by Quarter ($K)",
    data=[[120, 180, 210], [-80, -95, -110]],
    labels=["Q1", "Q2", "Q3"],
    series_names=["Revenue", "Expenses"],
    width=700, height=400,
).save("bar_sidebyside.svg")

# Column: multi-series with negatives
from charted.charts import ColumnChart

ColumnChart(
    title="Year-over-Year Growth Rate (%) by Segment",
    data=[[12, -8, 22, 18, -5, 30], [-3, -15, 5, -2, -20, 8], [9, -23, 17, 16, -25, 38]],
    labels=["Q1", "Q2", "Q3", "Q4", "Q5", "Q6"],
    width=700, height=500,
    theme={"v_padding": 0.12, "h_padding": 0.10},
).save("column.svg")

# Column: stacked (default for multi-series)
ColumnChart(
    title="Year-over-Year Growth by Segment",
    data=[[12, 22, 30], [-8, -15, -20], [4, 7, 10]],
    labels=["Q1", "Q2", "Q3"],
    series_names=["Revenue", "Costs", "Net"],
    width=700, height=400,
).save("column_stacked.svg")

# Column: side-by-side
ColumnChart(
    title="Sales Performance by Region",
    data=[[45, 52, 38, 61], [38, 46, 52, 49], [52, 39, 46, 51]],
    labels=["Q1", "Q2", "Q3", "Q4"],
    series_names=["North", "South", "East"],
    width=700, height=400, y_stacked=False,
).save("column_sidebyside.svg")

# Line: multi-series signal data
import math
from charted.charts import LineChart

n = 20
LineChart(
    title="Signal Analysis: Raw vs Filtered vs Baseline",
    data=[
        [math.sin(i * 0.5) * 30 + (i % 7 - 3) * 5 for i in range(n)],
        [math.sin(i * 0.5) * 25 for i in range(n)],
        [math.sin(i * 0.5) * 10 - 5 for i in range(n)],
    ],
    labels=[str(i) for i in range(n)],
    width=700, height=400,
).save("line.svg")

# Line: XY mode with temperature anomaly data
years = list(range(1990, 2010))
anomalies = [-15, -5, 10, 20, 5, 25, 15, 30, 10, 20, 40, 25, 45, 30, 50, 35, 60, 55, 45, 70]
baseline = [round(5 + 2 * math.sin(i * 0.4) + i * 0.5, 1) for i in range(len(years))]

LineChart(
    title="Temperature Anomaly vs 5-Year Rolling Baseline (1990-2009)",
    data=[anomalies, baseline],
    x_data=years,
    labels=[str(y) for y in years],
    width=700, height=400,
).save("xy_line.svg")

# Line: single series
LineChart(
    title="Monthly Active Users (K)",
    data=[[42, 48, 55, 61, 58, 70, 80, 78, 85, 92, 88, 100]],
    labels=["Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"],
    series_names=["MAU"], width=700, height=400,
).save("line_single.svg")

# Line: log scale (opt-in) for data spanning orders of magnitude.
# Pass y_scale="log" (or x_scale="log"); a LogScale instance also works.
LineChart(
    title="Requests/sec (log scale)",
    data=[[1, 8, 60, 450, 3200, 25000]],
    labels=["t0", "t1", "t2", "t3", "t4", "t5"],
    series_names=["rps"],
    y_scale="log",
    width=700, height=400,
).save("line_log.svg")

# Scatter: multi-series cluster analysis
import random
from charted.charts import ScatterChart

random.seed(42)
ca_x = [30 + random.gauss(0, 8) for _ in range(20)]
ca_y = [40 + random.gauss(0, 8) for _ in range(20)]
cb_x = [70 + random.gauss(0, 10) for _ in range(20)]
cb_y = [20 + random.gauss(0, 10) for _ in range(20)]

ScatterChart(
    title="Cluster Analysis: Two Distinct Populations",
    x_data=[ca_x, cb_x], y_data=[ca_y, cb_y],
    series_names=["Cluster A", "Cluster B"],
    width=700, height=400,
).save("scatter.svg")

# Scatter: single series with quadratic curve
random.seed(1)
x_vals = [i for i in range(5, 95, 5)]
y_vals = [round(10 + (v - 50) ** 2 / 50 + random.gauss(0, 4), 1) for v in x_vals]

ScatterChart(
    title="U-Shaped Response Curve: Signal vs Input",
    x_data=x_vals, y_data=y_vals,
    series_names=["Observations"],
    width=700, height=400,
).save("scatter_single.svg")

# Pie: basic
from charted.charts import PieChart

PieChart(
    title="Market Share by Product Line",
    data=[35, 28, 18, 12, 7],
    labels=["Product A", "Product B", "Product C", "Product D", "Other"],
    width=600, height=500,
).save("pie.svg")

# Pie: doughnut mode
PieChart(
    title="Operating System Market Share",
    data=[72, 15, 8, 5],
    labels=["Windows", "macOS", "Linux", "Other"],
    inner_radius=0.5, width=600, height=500,
).save("pie_doughnut.svg")

# Radar: multi-series
from charted.charts import RadarChart

RadarChart(
    title="Player Skill Comparison",
    data=[[85, 90, 75, 88, 92], [70, 85, 90, 75, 80]],
    labels=["Speed", "Strength", "Defense", "Technique", "Stamina"],
    width=600, height=500,
).save("radar.svg")

# Radar: single series
RadarChart(
    title="Character Stats",
    data=[20, 35, 30, 45, 25],
    labels=["Speed", "Power", "Endurance", "Defense", "Skill"],
    width=600, height=500,
).save("radar_multi.svg")

# Area: CPU temperature over 24 hours
from charted.charts import AreaChart

temps = [42 + 10 * math.sin(i * 0.6) + (hash(str(i)) % 5 - 2) * 1.5 for i in range(24)]

AreaChart(
    title="CPU Temperature (°C): 24-hour Cycle",
    data=[round(t, 1) for t in temps],
    labels=[f"{h}:00" for h in range(24)],
    width=700, height=400,
).save("area.svg")

# Area: multi-series revenue by channel
AreaChart(
    title="Multi-series Area: Revenue by Channel",
    data=[[30, 50, 45, 60, 70, 80, 65, 55], [20, 35, 30, 45, 50, 55, 40, 35]],
    labels=["Q1", "Q2", "Q3", "Q4", "Q5", "Q6", "Q7", "Q8"],
    series_names=["Online", "Retail"],
    width=700, height=400,
).save("area_multi.svg")

# Box Plot: distribution quartiles with outliers
import random
from charted.charts import BoxPlot

random.seed(42)
box_a = [round(random.gauss(50, 10), 1) for _ in range(50)] + [95, 5, 102]
box_b = [round(random.gauss(70, 15), 1) for _ in range(50)] + [120, 30, 130]
box_c = [round(random.gauss(30, 8), 1) for _ in range(50)] + [55, 8, 60]

BoxPlot(
    title="Test Scores by Group: with Outliers",
    data=[box_a, box_b, box_c],
    labels=["Group A", "Group B", "Group C"],
    width=700, height=400,
).save("boxplot.svg")

# Histogram: normal distribution (bell curve)
import random
from charted.charts import Histogram

random.seed(42)
scores = [random.gauss(50, 15) for _ in range(500)]

Histogram(
    title="Exam Scores: Normal Distribution (500 Students, 10 Bins)",
    data=scores,
    bins=10, width=700, height=400,
).save("histogram.svg")

# Heatmap: monthly temperature matrix
from charted.charts import HeatmapChart

HeatmapChart(
    title="Average Temperature (°C): Monthly by City",
    data=[
        [35, 36, 38, 40, 43, 45, 47, 46, 44, 41, 38, 36],
        [22, 24, 28, 32, 36, 40, 42, 41, 38, 33, 27, 23],
        [15, 18, 22, 27, 32, 37, 40, 39, 35, 29, 22, 17],
        [5, 8, 14, 20, 26, 32, 35, 34, 29, 22, 14, 7],
        [-2, 2, 10, 18, 25, 31, 34, 33, 27, 19, 10, 3],
    ],
    x_labels=["Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"],
    y_labels=["Dubai", "Sydney", "Tokyo", "Berlin", "Moscow"],
    width=700, height=450,
    low_color="#21639e", high_color="#f97316",
    show_values=True, value_format=".0f",
).save("heatmap.svg")

# Gantt: software project timeline
from charted.charts import GanttChart

GanttChart(
    title="Software Project Timeline: Q1 2026",
    data=[(0, 2), (1, 4), (3, 6), (5, 8), (6, 9)],
    labels=["Design", "Frontend", "Backend", "Testing", "Deployment"],
    width=700, height=400,
    dependencies=[(0, 1), (0, 2), (2, 3), (3, 4)],
    show_today_line=True,
    x_position=4.5,
).save("gantt.svg")

# Sankey: flow diagram (d3-sankey layout)
from charted.charts import SankeyChart

SankeyChart(
    title="Electricity generation to consumption",
    nodes=["Coal", "Gas", "Solar", "Grid", "Homes", "Industry", "Export"],
    links=[
        ("Coal", "Grid", 45), ("Gas", "Grid", 30), ("Solar", "Grid", 15),
        ("Grid", "Homes", 40), ("Grid", "Industry", 35), ("Grid", "Export", 15),
    ],
    width=800, height=500,
).save("sankey.svg")

Nodes are placed in columns by alignment, then relaxed over several iterations so
connected nodes line up and ribbons cross as little as possible. Link thickness is
proportional to flow value; the widths stacked at each node sum to the node's height.
Pass links as (source, target, value) tuples or {"source", "target", "value"}
dicts; endpoints may be node names or indices. Node labels in a dense column are
nudged apart vertically so they stay readable instead of overlapping.

The alignment option controls how nodes are assigned to columns: justify
(default, like d3-sankey) pushes every sink to the final column; left keeps each
node at its own depth; right aligns by distance from the sink; center is left
with sources pinned to the first column.

For funnel / drop-off data, use alignment="left". Funnels have dropout sinks
at many different depths (failed-filter, didn't-respond, cancelled, and so on). Under
the default justify those are all yanked into one tall final column, which crushes
the funnel shape. left lets each dropout terminate at its natural stage, so the
chart reads as a true funnel staircase:

# Funnel: recruitment drop-off, left-aligned so dropouts sit at their depth
SankeyChart(
    title="Recruitment funnel",
    nodes=[
        "Applied", "Passed screen", "Failed screen",
        "Interviewed", "Rejected", "Offered", "Declined", "Hired",
    ],
    links=[
        ("Applied", "Passed screen", 320), ("Applied", "Failed screen", 680),
        ("Passed screen", "Interviewed", 140), ("Passed screen", "Rejected", 180),
        ("Interviewed", "Offered", 60), ("Interviewed", "Rejected", 80),
        ("Offered", "Hired", 45), ("Offered", "Declined", 15),
    ],
    alignment="left",
    width=900, height=400,
).save("funnel.svg")

Theming

Three built-in presets (light, dark, high-contrast) plus custom theme composition:

from charted import BarChart

# Built-in themes
chart = BarChart(data=[120, 180, 210], labels=["Q1", "Q2", "Q3"], theme="light")
chart = BarChart(data=[120, 180, 210], labels=["Q1", "Q2", "Q3"], theme="dark")
chart = BarChart(data=[120, 180, 210], labels=["Q1", "Q2", "Q3"], theme="high-contrast")

Theme	Preview
Light
Dark
High Contrast

See the Theming docs for custom palettes, font overrides, and per-series styling.

CLI Usage

Generate charts without writing Python:

# From CSV
python -m charted create bar output.svg --data sales.csv

# From JSON
python -m charted create column chart.svg -d data.json

# Set the title and dimensions
python -m charted create bar output.svg --data sales.csv \
    --title "Q3 Sales" --width 900 --height 400

# Batch from directory
python -m charted batch input_data/ output_svg/

--title, --width, and --height override the same values in a --config
file when you pass both.

CSV format (default):

The first column is the x-axis labels. Every other column is a data series.

Quarter,Revenue,Expenses
Q1,120,80
Q2,180,95
Q3,210,110

Wide CSV (--transpose):

If your CSV is laid out sideways, with one series per row and the x values
across the header, pass --transpose. The corner cell is ignored, the rest of
the header row becomes the x-axis labels, and each following row is a series
named by its first cell. Without --transpose this layout would plot with the
axes swapped, so the flag is explicit rather than guessed.

Series,Q1,Q2,Q3
Revenue,120,180,210
Expenses,80,95,110

python -m charted create column out.svg --data wide.csv --transpose

JSON format:

{
  "labels": ["Q1", "Q2", "Q3"],
  "data": [[120, 180, 210], [80, 95, 110]],
  "series_names": ["Revenue", "Expenses"]
}

Full CLI docs: python -m charted --help

Data Loading

Load CSV/JSON without pandas:

from charted import load_csv, load_json, BarChart

# From CSV
x, y, labels = load_csv("sales.csv", x_col="Quarter", y_col="Revenue")
chart = BarChart(data=y, labels=x, title=labels[0])
chart.save("sales.svg")

# From JSON
x, y, labels = load_json("data.json")
chart = ColumnChart(data=y, labels=x)

Jupyter Notebook

Charts render inline automatically, no extra setup needed:

from charted.charts import BarChart

chart = BarChart(
    title="Sales by Quarter",
    data=[120, 180, 210, 150],
    labels=["Q1", "Q2", "Q3", "Q4"],
)
# Renders inline in the notebook cell

Markdown Export

from charted import BarChart

chart = BarChart(data=[120, 180, 210], labels=["Q1", "Q2", "Q3"], title="Sales")

# With file path
chart.save("docs/sales.svg")
md = chart.to_markdown(path="docs/sales.svg")  # ![Sales](docs/sales.svg)

# As inline data URL
md = chart.to_markdown()  # Data URL embedded in markdown

Base Chart Class

Dynamically select chart type at runtime:

from charted import Chart

chart = Chart(
    data=[120, 180, 210],
    labels=["Q1", "Q2", "Q3"],
    title="Sales",
    chart_type="bar",  # or column, line, scatter, pie, area, boxplot, histogram, heatmap, gantt
)
chart.save("chart.svg")

# Access all chart methods
svg = chart.to_svg()
md = chart.to_markdown()

Installation

Library (use in your code):

uv add charted

Or with pip:

pip install charted

CLI or MCP server (standalone tool, no project install needed):

uvx charted ...
# or
pipx install charted

Note: a uvx/pipx-installed package runs in its own isolated environment and cannot be imported into your project code. Use uv add or pip install when you want to import charted in your scripts.

Optional extras (these add dependencies, the core library stays zero-dep):

uv add "charted[png]"     # PNG export via cairosvg
uv add "charted[mcp]"     # MCP server for AI agent integration
uv add "charted[duckdb]"  # generate charts from SQL queries
uv add "charted[dev]"     # dev tools including PNG visual testing

# pip equivalents
pip install 'charted[png]'
pip install 'charted[mcp]'
pip install 'charted[duckdb]'
pip install 'charted[dev]'

PNG Export

Save charts directly as PNG by using the .png extension:

chart = BarChart(data=[10, 20, 30], labels=["A", "B", "C"])
chart.save("chart.svg")          # SVG (no extra dependencies)
chart.save("chart.png")          # PNG (requires cairosvg)
chart.save("chart.png", scale=3) # PNG at 3x resolution

PNG export requires cairosvg. If it's not installed, save() raises a helpful ImportError with install instructions.

MCP Server (AI Agent Integration)

Charted includes an MCP server so AI agents (Claude Code, Cursor, etc.) can generate charts without writing Python:

# Run standalone (no install step, recommended)
uvx --from 'charted[mcp]' charted-mcp

# Or install the extra and run directly
pip install 'charted[mcp]'
charted-mcp

# Register with Claude Code
claude mcp add charted -- uvx --from 'charted[mcp]' charted-mcp

Exposes tools: create_chart, list_chart_types, list_themes, chart_from_csv. The charted[mcp] extra is required.

create_chart and chart_from_csv take an output_format of svg, html, data_url, or png. With output_format="png" the tool returns a rasterized PNG image, so an agent can show the chart inline in a chat UI instead of relaying raw SVG markup. The mcp extra includes cairosvg, so PNG output works out of the box with the uvx command above. PNG rasterization needs cairo's system libraries; on Debian/Ubuntu install them with apt install libcairo2.

More features

A few extra parts of the public API that the examples above don't cover.

Auto chart selection. auto(data, **kwargs) picks a chart type from the shape of the data (1D becomes a bar or pie, a matrix becomes a heatmap, and so on) and returns a chart instance. auto_size(data, width, height) returns the (width, height) it would scale a dataset up to when you don't pass explicit dimensions.

from charted import auto

chart = auto([10, 20, 30], title="Sales")

Build from a dict or DataFrame. from_dict({"data": ..., "chart_type": ...}) builds a chart from a config dict. from_dataframe(df) takes a pandas DataFrame (or a plain dict of column to list if pandas isn't installed) and uses the first numeric column as the data and the index or first string column as the labels.

from charted import from_dict, from_dataframe

chart = from_dict({"chart_type": "line", "data": [1, 2, 3]})
chart = from_dataframe(df)  # falls back to a dict if pandas is missing

Inline and data-URL embedding. inline_svg(path) reads an SVG file back as a string for embedding in HTML or notebooks. chart_to_data_url(path) returns the same SVG URL-encoded as a data:image/svg+xml,... URI you can drop straight into an <img> tag or markdown image.

Save and restore a chart's config. chart.to_config() serializes a chart to a JSON-friendly dict (dimensions, data, labels, scales, reference lines, annotations, and so on). Chart.from_config(config, **overrides) rebuilds the chart from that dict, with keyword overrides merged on top so you can tweak one value without rebuilding the whole config.

config = chart.to_config()
chart2 = Chart.from_config(config, title="Updated title")

Fluent styling. chart.style(**kwargs) applies theme overrides and returns the chart for chaining, so you can set things like background_color or legend_font_size after construction.

chart = BarChart(data=[1, 2, 3]).style(background_color="#fff", legend_font_size=12)

Hover tooltips in HTML. chart.to_html(tooltips=True) attaches a native SVG <title> to each data mark so browsers show a built-in hover tooltip with no JavaScript. This only affects the HTML output; to_svg() and save() are unchanged.

Named palettes. resolve_palette(name) turns one of the built-in palette names into a list of hex colors you can pass as colors=. The names live in NAMED_PALETTES: default, viridis, ocean, categorical, rainbow, monochrome, pastel, sunset, forest, inferno, and the colourblind-safe okabe-ito.

from charted import resolve_palette

chart = BarChart(data=[1, 2, 3], colors=resolve_palette("viridis"))

Reference lines. Pass reference_lines=[{"value": 50, "axis": "y", "label": "Target"}] to draw a horizontal or vertical line at a value with an optional label. axis is "y" for a horizontal line (the default) or "x" for a vertical one.

Annotations. Pass annotations=[...] using LineAnnotation, BoxAnnotation, or LabelAnnotation to mark up the plot with lines, shaded regions, or text.

from charted import BarChart, LineAnnotation, BoxAnnotation, LabelAnnotation

Log and time scales. Pass x_scale= or y_scale= as "log" for a logarithmic axis or "time" for a time axis (which accepts dates, datetimes, or ISO date strings as x values). The default is "linear". Log and time scales are rejected on the value axis of a bar or column chart, since those fill from a zero baseline.

LineChart(data=[1, 10, 100, 1000], labels=["a", "b", "c", "d"], y_scale="log")

Links

Font System

Charted avoids tkinter by using pre-defined font metrics in fonts/definitions/. Generate new font definitions:

uv run python charted/commands/create_font_definition.py Helvetica