Machine readers · MCP

The Chart Color System as a tool.

The engine behind this app is available to AI assistants as a read-only Model Context Protocol server. An assistant that connects to it can solve a chart palette around your brand colors, audit colors you already have, show what a colorblind reader sees, and build sequential or diverging scales, with the engine's measurements and honesty notes instead of a guess.

Endpoint

https://chart-color-mcp.vercel.app/mcp

Streamable HTTP, stateless: JSON-RPC 2.0 over POST with Accept: application/json, text/event-stream. No authentication, no accounts, read-only. In Claude: Settings, Connectors, add custom connector, paste the URL. In Claude Code: claude mcp add --transport http chart-colors https://chart-color-mcp.vercel.app/mcp.

Tools

ToolWhat it returns
solve_paletteA categorical palette of n colors with your anchors kept exactly in slot order, the worst pair distances under normal and colorblind vision, the engine's relaxation notes, a full audit against your background, and a link that opens it in the app. Says so plainly above six slots, the honest cap.
audit_palettePass, warn or fail for colors you already have: pairwise OKLab separation under normal vision, deuteranopia, protanopia, tritanopia and achromatopsia, plus WCAG 2.2 non-text contrast of every color against the background.
simulate_paletteEach color as seen under each vision mode (published Machado 2009 matrices), so the assistant can describe or render the palette as a colorblind reader sees it.
build_rampAn evenly spaced perceptual scale, sequential or diverging, with a background-contrast audit.

The engine behind it

The same code the app runs, published as chart-color-system on npm. Every result reports the engine version. Distances are OKLab ΔE; contrast is WCAG 2.2.

Three things to ask

Privacy

The server keeps no accounts and stores nothing from a conversation. Requests are processed and discarded.

← Chart Color System