Context7
Up-to-date docs for AI agents
作者による説明(英語の原文)
Six pure, in-memory CSS color tools: parse, convert, WCAG/APCA contrast, gamut map, ramps, solver
この3つを1つのスコアにまとめることはしません。意見、測定、人気はそれぞれ別のものだからです。
このサーバーのやさしい解説はまだ書いていません。以下はレジストリから読み取った事実です。
使っているAIアプリを選んで、設定をコピーしてください。
必要なもの
どのアプリを使っていますか?
上でアプリを選んでください。選んだアプリを覚えておくので、次からはボタンに「…にインストール」と表示されます。
Claude Desktop → Settings → Developer → Edit Config を開き、claude_desktop_config.json にこれを貼り付けます
{
"mcpServers": {
"color-engine": {
"command": "npx",
"args": [
"-y",
"color-engine-mcp"
]
}
}
}ターミナルでこれを実行します
claude mcp add color-engine -- npx -y color-engine-mcpプロジェクトの .cursor/mcp.json、または ~/.cursor/mcp.json に貼り付けます
{
"mcpServers": {
"color-engine": {
"command": "npx",
"args": [
"-y",
"color-engine-mcp"
]
}
}
}ブラウザがアプリを開いてよいか確認し、そのあとアプリが設定を表示して確認を求めます。キーの値は空欄のままなので、自分で入力してください。
プロジェクトの .vscode/mcp.json に貼り付けます
{
"servers": {
"color-engine": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"color-engine-mcp"
]
}
}
}ブラウザがアプリを開いてよいか確認し、そのあとアプリが設定を表示して確認を求めます。キーの値は空欄のままなので、自分で入力してください。
これはパソコンで動くサーバーですが、ChatGPTがつなげるのはインターネット上のサーバーだけです。ほかのアプリを選んでください。
選んだアプリは、この端末のCookieに保存されます。アカウントは不要です。
この設定は、作者のレジストリ登録情報から作成しています。READMEと違う場合は、READMEに従ってください。
パソコンで動くサーバーです。インストールすると、ツール一覧がAIアプリに表示されます。
誰でも見られる情報をチェックしており、コードは読んでいません。当てはまらない項目は数に入れていません。 · 2026年11月7日にチェック
レジストリ名は、リポジトリを所有するGitHubアカウント aimsise で確認済みです
このリポジトリはまだGitHubから読み取っていません
このリポジトリはまだGitHubから読み取っていません
これはパソコンで動くため、インターネット上のエンドポイントはありません
パソコンで動き、代わりにローカルの設定を使います
パソコンで動くサーバーです。ツール一覧を読み取れるのは、インターネット上のエンドポイントがある場合だけです
レジストリでこのカテゴリにある有効なすべてのサーバーと比較しています。スターは、公開者がリポジトリの所有者である場合だけ数えます。
当てはまる基本チェック1項目のうち1項目に合格しました。コードは読んでいません。キーやトークンを追加する前に、サーバーがどんなアクセスを求めているかを確認し、必要な権限だけを持つキーを使ってください。
レジストリには必須のキーは載っていません。ただし、最初に使うときにサインインを求めるサーバーもあります。
あなたのパソコンで動きます。必要なときにAIアプリが起動し、データは作者のサーバーを通りません(サーバー自体がオンラインサービスを呼び出す場合を除きます)。
GitHubリポジトリの最終更新は–、最新バージョン(1.0.1)がレジストリに届いたのは5 か月前です。
Claude Desktop、Claude Code、Cursor、VS CodeなどのMCP対応アプリで使えます。パソコンにNode.jsも必要です。
同じカテゴリには:Context7, Codebase Memory, Exa
同じカテゴリで、ほかの作者が作ったサーバー。スターの多い順。
カードの見方:★ GitHubスターi3つの柱:レビュー · チェック · 利用状況i更新=最新バージョンをレジストリに公開した日i
Up-to-date docs for AI agents
Codebase knowledge graph for AI agents — 162 languages, sub-ms queries, 99% fewer tokens.
Connect AI agents to Exa for web search, content fetching, and multi-step research.
作者のリポジトリから書かれたとおりにコピーしたものです。PopNx MCPは内容を確認・推奨していません。HTML、画像、スクリプトは取り除いています。リンクはこのサイトの外へ移動します。
A Model Context Protocol (MCP) server providing 6 CSS color utilities: parsing, conversion, gamut mapping, WCAG contrast (with optional APCA), tint/shade ramp generation (with optional design-token output), and contrast-target solving. All tools operate purely in-memory — no network I/O, no filesystem writes.
| Tool | Purpose |
|---|---|
| parse_color | Parse any CSS color string into hex, RGB, OKLCH, and gamut info |
| convert_color | Convert a color into a canonical hex / rgb / hsl / oklch string |
| contrast | WCAG 2.1 contrast ratio and tier flags, with optional APCA Lc |
| gamut_map | Map wide-gamut colors into sRGB via perceptual OKLCH chroma reduction |
| generate_ramp | Tint-to-shade ramps with per-swatch contrast info and design-token output |
| solve_for_contrast | Find a foreground color that meets WCAG contrast targets against a background |
All JSON outputs shown in this README are genuine responses captured from the built server via the MCP Inspector CLI (
npx @modelcontextprotocol/inspector --cli).
Requires Node.js >= 20. The npm package exposes a color-engine-mcp bin, so no clone or build is needed — clients launch it via npx. This is a standard stdio MCP server using only the tools primitive, so any MCP-compatible client registers it the same way: point the client at the launch command, and the client spawns and manages the process itself.
For Claude Code, one command registers the server:
claude mcp add color-engine -- npx -y color-engine-mcp
(The default scope is the local project; add --scope user to register it for all your projects, or --scope project to write the shared .mcp.json.)
Most other JSON-configured clients accept this shape verbatim:
{
"mcpServers": { "color-engine": { "command": "npx", "args": ["-y", "color-engine-mcp"] } } }
Where the entry lives in a few common clients:
| Client | Config file |
|---|---|
| Claude Code | .mcp.json in the project root |
| Claude Desktop | claude_desktop_config.json (e.g. ~/Library/Application Support/Claude/ on macOS) |
| Cursor | .cursor/mcp.json (per-project) or ~/.cursor/mcp.json (global) |
| VS Code | .vscode/mcp.json — VS Code names the top-level key servers instead of mcpServers |
| Windsurf | ~/.codeium/windsurf/mcp_config.json |
Other clients (Zed, JetBrains, custom agents built on an MCP SDK, …) differ only in the file location and occasionally the top-level key name — see their MCP server registration docs.
Listed in the official MCP Registry as io.github.aimsise/color-engine-mcp.
Once connected, try asking your agent things like:
To run from a clone instead, see Development.
Every tool routes color strings through one shared parse boundary, so the rules below apply uniformly.
Accepted formats. Any CSS color string culori can parse, including named colors, hex (#rgb, #rgba, #rrggbb, #rrggbbaa), rgb()/rgba(), hsl()/hsla(), and oklch() — plus the wide-gamut / CSS Color 4 formats:
color(display-p3 r g b)lab(L a b) / lch(L C h)oklab(L a b)hwb(h w b)color(rec2020 r g b)color(a98-rgb r g b)color(xyz-d65 x y z)A wide-gamut input outside sRGB parses fine and reports inGamut: false:
{ "input": "color(display-p3 1 0 0)" }
{
"hex": "#ff0000", "rgb": { "r": 255, "g": 0, "b": 0 }, "oklch": { "l": 0.6485740751442981, "c": 0.2994852863383699, "h": 28.958132730803953 }, "inGamut": false }
Whitespace. Leading/trailing whitespace is trimmed before parsing — " #ff0000 " parses as #ff0000.
CSS Color 4 channel clamping (legacy spaces only). Out-of-range channels in hex / rgb() / hsl() inputs are clamped at the parse boundary, per CSS Color 4: rgb(-50 0 0) behaves exactly as rgb(0 0 0), and hsl() saturation/lightness clamp likewise (hue wraps). Inputs in other modes (oklch(), lab(), color(display-p3 …), …) are not clamped — their out-of-gamut values flow through raw, which is what makes gamut_map useful.
`none` channels. CSS Color 4 none channels are normalized to 0 in all six tools — oklch(0.5 none 30) behaves exactly as oklch(0.5 0 30).
Component magnitude. A parseable component with an absurd magnitude (above 1e6, e.g. oklch(0.5 1e30 30)) is rejected with COMPONENT_OUT_OF_RANGE — two tools (gamut_map, solve_for_contrast) surface it differently; see the error codes table. Real out-of-gamut values are many orders of magnitude below this guard and are never affected.
Alpha policy. contrast and solve_for_contrast reject translucent colors — any explicit alpha < 1, including rgba()/hsla() functional alpha and 4-/8-digit hex (#00000080) — with ALPHA_UNSUPPORTED, because the effective color of a translucent layer depends on an unknown backdrop; composite over the backdrop first. All other tools accept translucent input and simply ignore the alpha channel (computations use the opaque color; no output ever carries an alpha component).
Length cap. Color strings longer than 256 characters (after trimming) are rejected with INPUT_TOO_LONG.
Parse any CSS color string and return hex, RGB, OKLCH, and gamut info.
Input schema
| Field | Type | Required | Description |
|---|---|---|---|
| input | string | yes | Any CSS color string, e.g. "#ff0000", "red", "oklch(0.6 0.2 30)" |
Output for { "input": "#ff0000" }
{
"hex": "#ff0000", "rgb": { "r": 255, "g": 0, "b": 0 }, "oklch": { "l": 0.6279553639214311, "c": 0.2576833038053608, "h": 29.233880279627854 }, "inGamut": true }
The rgb channels are the sRGB-clamped 0–255 integer projection (consistent with hex). For an out-of-gamut input (e.g. a wide-gamut oklch(...)), the channels are clamped into [0, 255] rather than reporting raw out-of-range values — use the inGamut flag to detect that the input fell outside sRGB. The oklch block, by contrast, is the raw (lossless, unrounded) OKLCH of the input, and oklch.h is 0 for achromatic colors.
Convert a CSS color string into a canonical hex, rgb, hsl, or oklch format string.
Input schema
| Field | Type | Required | Description |
|---|---|---|---|
| input | string | yes | Any CSS color string |
| to | string | yes | Target format: "hex", "rgb", "hsl", or "oklch" |
Output for { "input": "#ff0000", "to": "oklch" }
{ "result": "oklch(0.62796 0.25768 29.23)" }
Raw vs. clamped for out-of-gamut inputs
For a color that lies outside the sRGB gamut, the target formats diverge:
to: "oklch" returns the raw, lossless OKLCH triple, so an out-of-gamut color round-trips faithfully. E.g. { "input": "oklch(0.7 0.4 30)", "to": "oklch" } → "oklch(0.70000 0.40000 30.00)".to: "hex", "rgb", and "hsl" are all derived from the sRGB-clamped projection and report the in-gamut approximation. E.g. { "input": "oklch(0.7 0.4 30)", "to": "rgb" } → "rgb(255, 0, 0)". rgb channels are integers in [0, 255].(L/C are formatted to 5 decimal places and H to 2, which guarantees an exact hex round-trip across the full sRGB cube.)
Compute the WCAG 2.1 contrast ratio between two fully opaque CSS color strings and return tier flags. Optionally also computes the APCA Lc value (see APCA).
Input schema
| Field | Type | Required | Description |
|---|---|---|---|
| a | string | yes | First CSS color string (treated as the text/foreground for APCA), e.g. "#000000" |
| b | string | yes | Second CSS color string (treated as the background for APCA), e.g. "#ffffff" |
| apca | boolean | no | When true, additionally return the signed APCA-W3 apcaLc for text a over background b |
Output for { "a": "#000000", "b": "#ffffff" }
{
"ratio": 21, "aaNormal": true, "aaLarge": true, "aaaNormal": true, "aaaLarge": true }
ratio is the 2-decimal display value. The four tier booleans are derived from the unrounded raw ratio (so a near-boundary raw 4.4999, which displays as 4.50, still yields aaNormal: false).
WCAG 2.1 tier thresholds:
aaNormal / aaLarge require ratio ≥ 4.5 / ≥ 3.0aaaNormal / aaaLarge require ratio ≥ 7.0 / ≥ 4.5Errors. A string that fails to parse yields a parameter-named error — PARSE_FAILED: could not parse the foreground color for a, PARSE_FAILED: could not parse the background color for b. Translucent input yields ALPHA_UNSUPPORTED (see Alpha policy).
Pass apca: true to additionally get apcaLc — the signed APCA-W3 (SAPC-4g) lightness contrast Lc, rounded to 2 decimals, for text `a` over background `b` (the argument order matters for APCA, unlike the symmetric WCAG ratio). The sign encodes polarity: positive for dark text on a light background, negative for light text on a dark background; compare magnitudes with |Lc|.
{ "a": "#1a1a1a", "b": "#ffffff", "apca": true }
{ "ratio": 17.4, "aaNormal": true, "aaLarge": true, "aaaNormal": true, "aaaLarge": true, "apcaLc": 104.27 }
Reversing the pair flips the sign: { "a": "#ffffff", "b": "#1a1a1a", "apca": true } → "apcaLc": -106.55 (same WCAG ratio of 17.4, since the WCAG ratio is symmetric).
Commonly cited APCA guideline thresholds (by |Lc|):
| \ | Lc\ | Common guideline use | |
|---|---|---|---|
| 45 | Minimum for large/bold text | ||
| 60 | Minimum for other content text | ||
| 75 | Body text | ||
| 90 | Preferred body text |
Disclaimer: APCA is a candidate method for WCAG 3 and is not yet a normative WCAG standard — use the WCAG 2.1 tier flags for conformance claims.
Map any CSS color string into the sRGB gamut via perceptual OKLCH chroma reduction. Useful for converting wide-gamut colors (P3, Rec2020, arbitrary OKLCH) to displayable sRGB.
Input schema
| Field | Type | Required | Description |
|---|---|---|---|
| input | string | yes | Any CSS color string, e.g. "oklch(0.6 0.4 30)" |
Output for { "input": "oklch(0.6 0.4 30)" }
{
"hex": "#f70000", "oklch": { "l": 0.6137129506655941, "c": 0.25183888824211537, "h": 29.233880279628004 }, "clamped": true }
clamped is true when the input was outside the sRGB gamut and had to be mapped. The returned oklch is the raw OKLCH of the mapped in-gamut result (chroma reduced perceptually), not of the original input.
Identity and idempotency. An input that is already inside the sRGB gamut is returned identically: clamped is false and hex is exactly the canonical hex of the input (e.g. { "input": "#3b82f6" } → { "hex": "#3b82f6", …, "clamped": false }). The tool is also idempotent: its own output hex is always in-gamut, so feeding a result back in takes the identity path and returns the identical hex.
G
冒頭だけを表示しています。続きはリポジトリで読めます