Skip to main content
Glama

Token-Efficient MCP

Cut your AI agent's token bill by ~92% on repetitive UI generation while guaranteeing brand consistency and schema validation.

Agents are great at logic, but terrible at writing HTML/CSS from scratch. Every time you ask an agent to "build a dashboard", it burns thousands of tokens writing boilerplate <table> markup, guessing hex codes, and inventing CSS layouts.

Token-Efficient MCP is a Model Context Protocol (MCP) server that gives agents a declarative UI language. Instead of generating code, the agent outputs a tiny JSON spec. The kit validates it against Zod schemas and renders pixel-perfect, responsive, brand-compliant HTML.

The Numbers

We benchmarked Claude Sonnet 5.5 (High effort) generating a variety of UI layouts.

Task

Raw HTML (Baseline)

JSON Spec (With Kit)

Reduction

Build a data table

2,533 tokens

134 tokens

94.7%

Build a KPI card

1,028 tokens

76 tokens

92.6%

Side-by-side layout

3,189 tokens

197 tokens

93.8%

Brand-constrained

2,986 tokens

177 tokens

94.1%

Complex multi-element

4,131 tokens

244 tokens

94.1%

Full dashboard grid

3,214 tokens

519 tokens

83.8%

Average token reduction: 92.2%

Read the full Benchmark Results.

Related MCP server: Dashboard MCP

Components

  • DataTable — Sortable, paginated table with sticky headers and empty state.

  • MetricCard — KPI card with trend indicators (up/down/flat).

  • ChartCard — Inline SVG bar/line charts with deterministic sorting and axis scaling.

  • DashboardGrid — 12-column responsive layout container. Place components side-by-side by assigning a span (1-12) to each child. Collapses to a single column on mobile.

How it works

  1. The agent calls list_components to see what's available.

  2. The agent outputs a JSON spec (e.g., { "component": "DataTable", "props": { "dataRef": "query_1" } }).

  3. The MCP server validates the spec against a strict Zod schema.

  4. The server renders the HTML, injecting brand tokens (colors, fonts, spacing) automatically.

The agent never writes CSS. It never hallucinates a hex code. It never inlines data.

Quick Start (Claude Desktop)

Add the following to your %APPDATA%\Claude\claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS).

No cloning or installation required—the server is fetched automatically via npx.

{
  "mcpServers": {
    "token-efficient-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@m4tic/mcp-server@0.1.3"
      ]
    }
  }
}

Restart Claude Desktop, and the dashboard-kit tools will be available.

(Note: By default, the server looks for brand configs in your local examples/brand-configs/ folder. If you are using it via npx, it will gracefully fall back to the default light or dark theme if custom brand files are not found).

Revenue fixtures load first from examples/fixtures/revenue.json relative to the client's working directory, then from the server package's bundled dist/fixtures/revenue.json. Bundled fixture use is logged to stderr. If neither file exists, tables render their empty state.

Layout defaults

The HTML shell fills the browser viewport. DashboardGrid automatically packs children into 12 columns. When there is more than one row, a first row containing only metrics fits its content; every other row shares the remaining height. A single row fills the available height. Metric spans above 4 are clamped to 4. On desktop, metric cards stop growing at 180px and table cards at 700px. Tables scroll internally; charts retain readable labels with internal scrolling when space is tight. On mobile, cards stack and height caps are removed.

Agents can omit height and rows. Explicit rows overrides the automatic tracks; legacy height values are shrinkable desktop size preferences. Number and currency columns default to right alignment; explicit alignment still wins. Trend values are optional. A flat metric trend without a value displays ■ Flat; a supplied value, including zero, displays its percentage.

Features

  • Brand Enforcement: Injects brand tokens via CSS variables at render time. Zero hard-coded hex codes.

  • Light/Dark Mode: Supports full brand inheritance. Pass just "mode": "dark" and the server builds a coherent dark theme.

  • Zod Validation: Rejects malformed specs before they hit the UI.

  • Token Budgets: Each component declares its max allowed token cost.

Built by M4TIC

We build intelligent systems, AI agents, and workflow automations. If you need to connect your company's design system to the agent stack, get in touch.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      B
      maintenance
      Enables AI agents to create and manage customizable dashboards with text, list, stats, progress, and chart widgets, accessible via a read-only web viewer and JSON API.
      -
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables AI agents to run deterministic API validation and UI DOM checks, verifying HTTP status codes, JSON schema fields, and page elements with programmatic proof instead of LLM guesswork.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      It lets AI coding agents look up design tokens, component specifications, and prohibition rules on demand, then validate generated HTML or JSX with the same lint logic used by CI and editor hooks so violations are caught and corrected immediately. It also exposes the design constitution and rule sets as resources for reference during generation.
      6
      206 npm
      202
      MIT