io.github.dockster/uxspot-mcp

uxspot MCP server

The UX reference your AI assistant can call — and now, the one that draws.

uxspot.io is a free, curated UX reference: a 203-term glossary (every term with a plain-English definition and a "why it matters"), an AI x UX tools directory, 5 interactive checklists, a 78-topic Learn UX curriculum, and Spot Check, a daily UX vocabulary game. This repository documents the remote Model Context Protocol server that exposes all of it as callable tools for Claude, Cursor, and any other MCP client.

• Endpoint: https://uxspot.io/mcp — remote, Streamable HTTP, stateless
• Landing page and live demo: uxspot.io/mcp-server
• Auth: none — no key, no account, no rate-limit signup
• Privacy: every tool is read-only over public uxspot.io content. The server never sees your project, files, or account.
• Sibling project: the uxspot Figma plugin — the same reference as a design mentor inside Figma, live on Figma Community.

Install

Claude Code

claude mcp add --transport http uxspot https://uxspot.io/mcp

Claude Desktop — Settings → Connectors → Add custom connector, with URL https://uxspot.io/mcp.

Cursor — add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "uxspot": { "url": "https://uxspot.io/mcp" }
  }
}

Gemini app (Spark) — Settings → Connected apps → add a custom app, and paste https://uxspot.io/mcp as the URL (just the URL — the JSON below is Antigravity's format, not Spark's). If a credentials step appears under Advanced, leave it empty; uxspot never asks for a key.

Gemini (Antigravity) — the IDE and CLI that replaced Gemini CLI. Add to ~/.gemini/config/mcp_config.json, then /mcp in the CLI to confirm:

{
  "mcpServers": {
    "uxspot": { "serverUrl": "https://uxspot.io/mcp" }
  }
}

Gemini Enterprise is the one Gemini door that stays closed: it requires an admin-registered OAuth connector, and uxspot is keyless by design.

Any other MCP client that supports the Streamable HTTP transport: point it at https://uxspot.io/mcp.

Tools

Every answer cites its uxspot.io source URL, so you (and your assistant) can verify it.

What an answer looks like

define_term("60-30-10") returns, verbatim:

**60-30-10 rule**
A color-proportion guideline borrowed from interior design: roughly 60% of a
screen in a dominant, usually neutral color, 30% in a secondary color, and 10%
in an accent reserved for emphasis — enough color to guide the eye without
competing for it.

**Why it matters** — Screens that feel colorful but calm almost always follow
some version of this split. Inverting it — letting the accent become the
majority — is why loud interfaces feel loud: when everything asks for
attention with color, nothing gets it.

**The numbers**
- 60% — dominant, usually neutral
- 30% — secondary, supporting surfaces
- 10% — accent, reserved for emphasis and actions

┌────────────────────────────────────┐
│                                    │
│      60% — dominant, neutral       │
│                                    │
├─────────────────────────┬──────────┤
│    30% — secondary      │ 10%      │
│                         │ accent   │
└─────────────────────────┴──────────┘

Source: https://uxspot.io/glossary/60-30-10-rule

The terminal diagrams

37 spatial concepts — proportion, position, hierarchy, flow — answer with hand-drawn box art: the Z-pattern's scan path, the information-architecture tree with its misfiled page, the journey map's emotional dip, the empathy map's quadrants, the modal punched through a dimmed page.

The set is deliberately bounded. Terms whose specimen needs color (contrast ratios, eye-tracking heat), motion (skeleton shimmer), or interaction (Fitts's law's timer, the Zeigarnik checklist) don't get a worse imitation — they point to their live interactive example on uxspot.io instead. And some things monospace physically cannot show: a kerning diagram in a terminal would be a lie about its own subject, since every glyph is the same width by definition.

Every diagram is hand-authored, never generated, in src/lib/glossaryAscii.ts — inspectable like everything else here, and the file records the full adjudication of what is drawn, what points to a live demo, and why.

Try it

Ask your assistant things like:

• "What's the Z-pattern? Draw it."
• "What's an affordance in UX — and where can I read more?"
• "List AI tools for UX research"
• "Give me the accessibility checklist"
• "Search the curriculum for usability testing"
• "Quiz me on UX vocabulary" (runs Spot Check, round by round)

How it is built

The server is a single Cloudflare Pages Function in front of the static uxspot.io site. Each POST is a complete JSON-RPC exchange (stateless Streamable HTTP — no sessions, no database), and the tools read the same structured data that powers the live site, so responses are always in sync with what is published. The tool descriptions even self-count from the data files, so the numbers above cannot drift.

The code in server/ and the diagram map in src/lib/ are a read-only mirror of the deployed function, published here for transparency and for MCP directory listings. The source of truth lives in the main uxspot site repository; the data files the function imports (glossary.json, tools.json, checklists.json, curriculum.ts) live there too, so this mirror is for reading, not building.

License

The server code in this repository is released under the MIT License. The uxspot.io content the tools serve (definitions, checklists, curriculum) remains © Soufiane Chraibi and is licensed for use via the site and this API, not for bulk redistribution.