[FEATURE] Multi-dimensional / scientific data visualization in Claude Code (charts, plots, heatmaps — beyond Mermaid)

Preflight Checklist

• [x] I have searched existing requests. This is distinct from:
• #14375 / #52517 / #20529 — Mermaid (structural node-and-edge diagrams)
• #13600 — markdown rendering
• #64643 — PostAssistantMessage hook + OSC1337/Sixel passthrough (the transport)
• #60810 — view_image (image input)
• [x] This is a single feature request.

Problem Statement

The existing Mermaid requests cover structural diagrams (nodes + edges).
But a large class of visualization needs — data analysis, monitoring, trading,
profiling — require analytical charts that Mermaid structurally cannot
express:

• multi-variable relationships: 4D (x / y / size / color), parallel coordinates
• distributions: histograms, box plots, heatmaps, violin plots
• time series with overlays (price + volume + indicators)
• 3D surfaces / volumetric data

Claude Code currently has zero path for any of these:

• terminal is text-only
• ASCII bar charts break with multiple series and can't do scatter/heatmap/3D
• no Vega-Lite / plotly / echarts rendering
• "emit HTML/PNG + open" breaks conversation flow and isn't replayable

claude.ai web already renders a broader set of these; the CLI currently
renders none. The gap is most painful for workflows that live entirely in
the terminal.

Proposed Solution

Two layers, both already half-built in the community:

1. Transport (#64643): PostAssistantMessage hook + OSC 1337 / Sixel /

Kitty passthrough → rendered pixels land inline in capable terminals.
This issue requires that transport carry rendered chart images, not
just Mermaid output.

1. Render layer: a sandboxed renderer for a declarative chart spec.

Vega-Lite is the right pick — JSON (auditable in transcript, diff-able),
declarative, covers the entire analytical chart space Mermaid can't.
Claude emits the spec; CC renders to SVG/PNG and pipes through the transport.

Use Cases

• blockchain: gas distribution histograms, MEV/swap scatter, correlation

heatmaps across pools, 4D (price x volume x slippage x time)

• ops/SRE: timeseries dashboards, latency heatmaps, flame graphs
• data science: the entire EDA workflow

Priority

High — blocks Claude Code for any analytical / quant / ops workflow.

Why not duplicate

Sibling issues solve "how do we draw structure." This solves "how do we
draw data." Same transport (#64643), different render layer. Land the hook
once, then Mermaid (#14375) and analytical charts (this) ride the same pipe.