Verified conversions across 205 color spaces: CSS strings, gamut checks, .cube LUTs
Runs on your machine from its package; the steps are below.
At most: only reads or looks up information.
Odel charges nothing per call. The publisher may charge for its own service.
Runs locally over stdio
This server isn't hosted — your MCP client launches it from a package registry. Use one of the commands below, or drop the config into your client (e.g. Claude Desktop).
Run
npx -y color-space
MCP client config
{
"mcpServers": {
"color-space": {
"command": "npx",
"args": [
"-y",
"color-space"
]
}
}
}Open collection of color spaces.
import space from 'color-space';
space.rgb.oklch(255, 128, 0); // → [0.732, 0.186, 53] — modern CSS
space.slog3.rec2020(0.5, 0.5, 0.5); // camera log → UHD wide gamut
space.rgb.xyz(300, -10, 0); // → [59.7, 30.6, 2.8] — out-of-range in, unclamped out
// batch
space.rgb.oklch([255, 128, 0, 0, 255, 128]); // [0.732, 0.186, 53, 0.875, 0.234, 151]
// meta
space.lab.range; // [[0, 100], [-125, 125], [-125, 125]]
Individual spaces:
import oklch from 'color-space/oklch.js';
oklch.rgb(0.65, 0.25, 180); // matches CSS oklch(0.65 0.25 180)
color-space – All 205 interconnected spaces · 55 kB gzipcolor-space/<name>.js – One standalone spacecolor-space/lite – 54-space working set · 29 kB gzipcolor-space/wasm – lite set as WASM · bare .wasm includedcolor-space/gl – GLSL/WGSL shader sourcecolor-space/gl/hlsl · color-space/gl/msl – HLSL (DirectX, Unity, Unreal) and Metal, written by naga from the WGSL · await hlsl('rgb', 'oklch') · needs the optional peer naga-wasm (it declares Node ≥ 24.12; CI runs it on 22, where npm only warns)color-space/lut – .cube LUTs for Resolve, Premiere, OBS, ffmpeg — verified vs ACEScolor-space/icc – Matrix + TRC or CLUT profilescolor-space/data.json – Channels, ranges, provenance, references, graph, gamutsnpx color-space rgb oklch 255 128 0 – CLI, no install: convert (default) · cube · icc · space · spaces · mcp (MCP server for agents)MCP tools – convert · gamut · css · space · spaces · cube · shader · icc · code – so agents call verified math instead of guessing it. A color goes in as any CSS string ('#ff8000', 'oklch(0.7 0.1 50)', 'rebeccapurple') or as channel values; shader writes the conversion as GLSL, WGSL, HLSL or MSL, code gives the JS call and its verified colour-science equivalent.
claude mcp add color-space -- npx -y color-space mcp # Claude Code
code --add-mcp '{"name":"color-space","command":"npx","args":["-y","color-space","mcp"]}' # VS Code
Claude Desktop claude_desktop_config.json, any project .mcp.json:
{ "mcpServers": { "color-space": { "command": "npx", "args": ["-y", "color-space", "mcp"] } } }
Claude Code plugin (the server plus a skill): /plugin marketplace add colorjs/color-space, then /plugin install color-space@color-space. Docker: docker build -t color-space-mcp . and run it with docker run -i --rm color-space-mcp.
Over HTTP (Streamable HTTP, stateless): npx -y color-space mcp --http 8787 on localhost, or host fetch from color-space/mcp.js in a Worker, Deno or Bun. Skill for agents that load skills: plugins/claude/skills/color-space.
A complete collection of color spaces: one minimal API, verified formulas. While alternatives focus on digital color spaces, this project takes broader perspective, covering historical and cross-disciplinary spaces as well. Side effects: corrected papers, color education, test cases for JS→WASM compilers (porffor, jz).
It is not a color toolbox — parsing, interpolation, ΔE, gamut mapping, contrast, palettes, alpha stay out. Pantone, NCS, RAL Classic, and similar licensed swatch catalogs aren't open and stay out as well.
Thanks to the researchers, theorists, specifiers, implementors and the libraries that helped this one happen.
| Library | Spaces | Ranges | Camera log · CAM · historic | Backends | Speed (M op/s)¹ |
|---|---|---|---|---|---|
| color-space | 205 | Conventional | ✅ | JS · WASM · GLSL/WGSL · LUT · ICC | 5.52 |
| color-space/lite | 54 | Conventional | logs · HDR · CAM16 | JS · 29 kB gzip | 3.01 ² |
| color-space/wasm | 54 | Conventional | logs · HDR · CAM16 | WASM | 1.38–4.78× JS batches ³ |
| culori | 30 | 0–1 | ❌ | JS | 4.46 |
| colorjs.io | ~40 | 0–1 | some | JS | 0.23 |
| texel/color | ~16 | 0–1 | ❌ | JS | 3.49 ² |
| chroma-js | ~12 | mixed | ❌ | JS | 1.19 ² |
| d3-color | 6 | mixed | ❌ | JS | 9.35 ² |
¹ geometric mean, million scalar calls per second over the 7 shared npm run benchmark conversions (rgb ⇄ lab · hsl · oklab, rgb → p3), inputs rotating through a 256-color table; median of 3 runs on 2026-10-03 as the benchmark prints it, Node 22.22 / Intel Xeon 2.1 GHz (4 vCPU), culori 4.0.2 · colorjs.io 0.5.2 · @texel/color 1.1.11 · chroma-js 3.2.0 · d3-color 3.1.0. Absolute figures move with the machine; the ratios are the claim · ² measured on the subset it implements — color-space on the same subset: lite's 4 (lab, oklab; measured before lite gained hsl and p3) 3.09 · texel's 3 (oklab, p3) 2.76 · chroma-js's 6 6.45 · d3-color's 4 (lab, hsl) 9.28 · ³ WASM speedup over the JS batch API, rgb → each of the 26 targets it had then over 1M pixels (the 27 added since are not yet measured), per-target medians of 3 further runs the same day (lowest on Jzazbz and JzCzHz)