Your first chart
A guided walkthrough of authoring, binding, and exporting a chart.
Pick the data
Charts bind to a dataset. If you followed the quickstart you already have one; otherwise capture one from a connection first.
Choose a family
Charts are declarative: you describe what to show, not how to draw it. Supported families include bar, line, area, scatter, bubble, pie, donut, histogram, heatmap, table, KPI, boxplot, candlestick, waterfall, funnel, radar, gauge, treemap, sunburst, Sankey, graph, choropleth, geographic point, and small multiples.
Each family declares which field channels it supports and what field types they accept. The platform validates your mapping against the dataset's fields before rendering — mismatched types, missing measures, or oversized point counts are rejected with actionable errors.
Map fields to channels
Every family defines a set of channels. A typical bar chart uses:
| Channel | Example | Notes |
|---|---|---|
| Category | region | Dimension field |
| Value | revenue | Measure field |
| Color | — | Optional series break |
The editor exposes the common controls; the full declarative surface is also available through the API, MCP, and SDK, so anything the console can do, an agent can do too.
Bind live or snapshot
A chart references its data through a binding:
- Live reflects the dataset as it changes — with explicit manual refresh or an authorized schedule.
- Snapshot pins the chart to one frozen capture, so a chart you shared never changes underneath its viewers.
See chart bindings for revision semantics.
Export
Render the chart to PNG or SVG, or export its data as snapshot JSON or CSV. Exports are queued jobs with immutable manifests — you always know which revision and capture an output came from. See exports.