thevangelist/dembrandt

Dembrandt.

Extract a website's design system into design tokens in a few seconds: logo, colors, typography, borders, and more. One command.

Install

Install globally: npm install -g dembrandt

dembrandt dembrandt.com

Or use npx without installing: npx dembrandt dembrandt.com

Requires Node.js 18+

AI Agent Integration (MCP)

Use Dembrandt as a tool in Claude Code, Cursor, Windsurf, or any MCP-compatible client. Ask your agent to "extract the color palette from dembrandt.com" and it calls Dembrandt automatically.

claude mcp add --transport stdio dembrandt -- npx -y --package dembrandt dembrandt-mcp

Or add to your project's .mcp.json:

{
  "mcpServers": {
    "dembrandt": {
      "command": "npx",
      "args": ["-y", "--package", "dembrandt", "dembrandt-mcp"]
    }
  }
}

Available tools include get_design_tokens, get_color_palette, get_typography, get_component_styles, get_surfaces, get_spacing, and get_brand_identity, plus pure analysis tools (compute_drift, get_findings, export_dtcg, generate_design_md, render_report) and job-control tools. Extraction tools accept mobile, cookie (for authenticated pages), and wcag options.

Pair with dembrandt-skills to give your agent UX intelligence on top of extracted tokens — hierarchy, accessibility, interaction states, and a full 6-stage design pipeline orchestrator.

npx skills add dembrandt/dembrandt-skills

Dembrandt App (Beta)

Load extractions, track token drift, and compare snapshots. dembrandt.com/app

• Automatic drift tracking from CI. Generate an API key at dembrandt.com/app/api-keys, then pass --key to the CLI. Every run uploads a snapshot to your account and scores it against the previous one for that domain. Wire into GitHub Actions or any CI runner and every deploy records itself.
• Pin a baseline. Mark any snapshot as your reference. Every subsequent extraction is automatically scored against it.
• Visual diff. Color swatches, before/after values, delta scores per category: colors, typography, spacing, radius, shadows.
• Snapshot timeline. Proportional timeline per domain — scrub across any date range from days to years.
• Compare side by side. Load multiple extractions into one view: two releases, two sites, or two surfaces.
• Copy tokens. Paste values straight into Copilot, Claude, or Cursor.
• No login required for local use. Data stays in the browser. Sign in with GitHub to enable cloud sync.

Recipes

dembrandt.com/recipes — ready-to-run workflows. Copy a command, paste a prompt, get a result. Covers competitor benchmarking, WCAG audits, CI/CD drift detection, Figma token push, and agentic design system builds. Filterable by role.

What to expect from extraction?

• Colors (semantic, palette, CSS variables, gradients)
• Typography (fonts, sizes, weights, sources)
• Spacing (margin/padding scales)
• Borders (radius, widths, styles, colors)
• Shadows
• Motion (duration scale, easing curves, hover patterns per component type)
• Components (buttons, badges, inputs, links)
• Breakpoints
• Icons & frameworks

Usage

dembrandt <url>                        # Basic extraction (terminal display only)
dembrandt dembrandt.com --json-only     # Output raw JSON to terminal (no formatted display, no file save)
dembrandt dembrandt.com --save-output   # Save JSON to output/dembrandt.com/YYYY-MM-DDTHH-MM-SS.json
dembrandt dembrandt.com --dtcg          # Export in W3C Design Tokens (DTCG) format (auto-saves as .tokens.json)
dembrandt dembrandt.com --dark-mode     # Extract colors from dark mode variant
dembrandt dembrandt.com --mobile        # Use mobile viewport (390x844) for responsive analysis
dembrandt dembrandt.com --slow          # 3x longer timeouts (24s hydration) for JavaScript-heavy sites
dembrandt dembrandt.com --brand-guide   # Generate a brand guide PDF
dembrandt dembrandt.com --design-md     # Generate a DESIGN.md file for AI agents
dembrandt dembrandt.com /pricing /docs  # Extract specific paths and merge results into one output
dembrandt dembrandt.com --crawl 5       # Analyze 5 pages (homepage + 4 discovered pages), merges results
dembrandt dembrandt.com --sitemap       # Discover pages from sitemap.xml instead of DOM links
dembrandt dembrandt.com --crawl 10 --sitemap # Combine: up to 10 pages discovered via sitemap
dembrandt dembrandt.com --no-sandbox    # Disable Chromium sandbox (required for Docker/CI)
dembrandt dembrandt.com --key dmb_···  # Push snapshot to your Dembrandt account; auto-scored against the previous snapshot for that domain
                                       # DEMBRANDT_API_URL env var overrides the upload endpoint (default: https://www.dembrandt.com)
dembrandt dembrandt.com --browser=firefox # Use Firefox instead of Chromium (better for Cloudflare bypass)
dembrandt dembrandt.com --wcag          # WCAG 2.1 contrast analysis — real DOM pairs, AA/AAA grades
dembrandt dembrandt.com --stealth       # Opt-in anti-detection: navigator spoofing + human mouse simulation (use only when authorized)
dembrandt dembrandt.com --locale fi-FI --timezone Europe/Helsinki # Browser fingerprint: locale and timezone
dembrandt dembrandt.com --user-agent "Mozilla/5.0 ..."           # Custom user agent string
dembrandt dembrandt.com --accept-language "fi,en;q=0.9"          # Custom Accept-Language header
dembrandt dembrandt.com --screen-size 2560x1440                  # Physical screen resolution to report

Default: formatted terminal display only. Use --save-output to persist results as JSON files. Browser automatically retries in visible mode if headless extraction fails.

All flags combine unless noted otherwise: see docs/FLAGS.md for the flag compatibility tables (interactions, ignored combinations, multi-page propagation).

Multi-Page Extraction

Analyze multiple pages to get a more complete picture of a site's design system. Results are merged into a single unified output with cross-page confidence boosting: tokens appearing on multiple pages get higher confidence scores.

# Analyze homepage + 4 auto-discovered pages (default: 5 total)
dembrandt dembrandt.com --crawl 5

# Use sitemap.xml for page discovery instead of DOM link scraping
dembrandt dembrandt.com --sitemap

# Combine both: up to 10 pages from sitemap
dembrandt dembrandt.com --crawl 10 --sitemap

Page discovery works two ways:

• DOM links (default): Reads navigation, header, and footer links from the homepage, prioritizing key pages like /pricing, /about, /features
• Sitemap (--sitemap): Parses sitemap.xml (checks robots.txt first), follows sitemapindex references, and scores URLs by importance

Pages are fetched sequentially with polite delays. Failed pages are skipped without aborting the run.

Browser Selection

By default, dembrandt uses Chromium. If you encounter bot detection or timeouts (especially on sites behind Cloudflare), try Firefox which is often more successful at bypassing these protections:

# Use Firefox instead of Chromium
dembrandt dembrandt.com --browser=firefox

# Combine with other flags
dembrandt dembrandt.com --browser=firefox --save-output --dtcg

When to use Firefox:

• Sites behind Cloudflare or other bot detection systems
• Timeout issues on heavily protected sites
• WSL environments where headless Chromium may struggle

Installation: Browsers are installed on demand, not by npm install (dembrandt depends on the lean playwright-core, which carries no browser binaries). Fetch the engine you need, matched to the installed playwright-core:

dembrandt install-browser           # chromium (default)
dembrandt install-browser firefox   # a specific engine

This resolves the browser revision from the playwright-core dembrandt actually drives, so the two cannot drift apart. Prefer it over calling Playwright directly: a bare npx playwright install fetches whatever version the registry serves, and a mismatch fails with Executable doesn't exist.

On Linux and in CI, system libraries are installed separately:

npx playwright@$(node -p "require('playwright-core/package.json').version") install --with-deps chromium

Connect to an existing browser (CDP)

Skip the bundled browser entirely and drive an already-running Chromium over the DevTools Protocol. Useful in CI or containers where a browser is already up, and it needs no local browser download at all:

BROWSER_CDP_ENDPOINT=http://localhost:9222 dembrandt dembrandt.com --browser chromium

CDP is supported only with --browser chromium.

W3C Design Tokens (DTCG) Format

Use --dtcg to export in the standardized W3C Design Tokens Community Group format:

dembrandt dembrandt.com --dtcg
# Saves to: output/dembrandt.com/TIMESTAMP.tokens.json

The DTCG format is an industry-standard JSON schema that can be consumed by design tools and token transformation libraries like Style Dictionary.

DESIGN.md

Use --design-md to generate a DESIGN.md file, a plain-text design system document readable by AI agents. The export follows Google's DESIGN.md draft format: YAML design tokens in front matter plus ordered Markdown guidance sections.

dembrandt dembrandt.com --design-md
# Saves to: output/dembrandt.com/DESIGN.md

DESIGN.md reports only what Dembrandt observed on the source site. Exact values (colors, typography, spacing, radii, shadows) live in the YAML front matter when available, and the Markdown body adds human-readable context. Sections with no extracted evidence are omitted rather than filled with invented defaults. For example, the elevation section is dropped when the site uses no box-shadow tokens.

WCAG Contrast Analysis

Use --wcag to check accessibility contrast ratios across the page. Unlike palette-based checkers, dembrandt walks the actual DOM and finds what color is rendered on top of what background — per element.

dembrandt dembrandt.com --wcag

Returns every text/background pair with contrast ratio and WCAG 2.1 grade (AA, AA-Large, AAA, or fail), sorted by how often each pair appears. Results are shown in terminal and included in JSON output as wcag.

Also captures interactive state contrast: dembrandt simulates hover, focus, and disabled states on buttons, links, and inputs and checks contrast on each state. State pairs are tagged [hover], [focus], or [disabled] in output so you can catch contrast failures that only appear on interaction.

Motion Tokens

Motion tokens are extracted automatically on every run — no flag needed. Dembrandt analyzes CSS transitions and animations across the page and returns a structured motion profile.

dembrandt dembrandt.com

Returns:

• Duration scale: all unique animation durations found on the page
• Easing curves: named easing types (ease-out, spring, custom cubic-bezier) with usage counts
• Per-context profiles: motion behavior by component type (button, nav, card, modal, hero)
• Hover interaction deltas: which properties animate on hover (transform, opacity, background, color) and the pattern (scale-up, fade-in, color-shift, slide-y)

Motion data is included in JSON output as motion and printed in terminal under a dedicated Motion section.

ML-powered brand color detection (experimental)

dembrandt dembrandt.com --ai
#   ⚡ ML primary → #533afd (score 0.93 · 68% acc)

Replaces the heuristic with a trained model — 2× more accurate (68% vs 32%). Requires the optional onnxruntime-node dep (npm install onnxruntime-node). Without the flag nothing changes.

Brand Guide PDF

Use --brand-guide to generate a printable PDF summarizing the extracted design system: colors, typography, components, and logo on a single document.

dembrandt dembrandt.com --brand-guide
# Saves to: output/dembrandt.com/TIMESTAMP.brand-guide.pdf

Continuous integration

GitHub Action

The official action wraps extract → compare → gate into one step: it installs a matching Chromium, runs a pinned CLI version, fails the job on drift, and renders the drifted tokens as inline annotations on the PR.

- uses: dembrandt/[email protected]
  with:
    url: https://preview.example.com
    baseline: .dembrandt/baseline.json

Gating a Vercel preview needs no extra wiring — trigger on the deployment event and pass its URL:

on:
  deployment_status:

jobs:
  drift:
    if: github.event.deployment_status.state == 'success'
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: dembrandt/[email protected]
        with:
          url: ${{ github.event.deployment_status.environment_url }}
          baseline: .dembrandt/baseline.json
          key: ${{ secrets.DEMBRANDT_KEY }}

Outputs: report (path to the drift/extraction JSON) and score (drift score, empty without a baseline). The action pins the CLI version per release, so a version-tagged gate never changes behavior under you. For a fully hand-rolled workflow (per-page PR comment, preview vs production, report artifact), see examples/drift-gate.yml.

Running the CLI directly

Dembrandt drives a real browser, so the browser revision must match playwright-core.

If you are not using the Playwright container image, install the browser revision that matches playwright-core:

# matches the bundled playwright-core automatically
dembrandt install-browser
# on a bare Linux runner, add the system libraries too
npx playwright@$(node -p "require('playwright-core/package.json').version") install --with-deps chromium

A mismatched version fails with "Executable doesn't exist". The container image avoids this entirely — just match its tag (v1.60.0) to the playwright-core version.

Drift gate

Compare an extraction against a committed baseline and fail the job on drift:

# capture a baseline once (same environment you will check against)
dembrandt https://app.example.com --json-only > baseline.json

# in CI — exits non-zero on drift; writes a report artifact
dembrandt https://app.example.com --compare baseline.json --html report.html

When the change is intended, accept it as the new baseline — --approve overwrites the local baseline file and passes instead of failing:

dembrandt https://app.example.com --compare baseline.json --approve

Add --json-only to a --compare run to get the drift report as machine-readable JSON under a drift key — score, status, summary, and per-token changes[] (each with category, kind, before, after, delta). A CI gate can render exactly which tokens moved (e.g. in a PR comment) from this instead of parsing the HTML report:

dembrandt https://app.example.com --compare baseline.json --json-only

Any CI. The gate is platform-neutral — it is just the exit code plus the drift JSON, so it drops into any runner:

dembrandt "$PREVIEW/checkout" --compare base.json --json-only > drift.json
# exit 1 = drift. Read drift.json (.drift.changes) and surface it however your
# platform does: a GitLab MR note, an Azure DevOps PR thread, a Jenkins status,
# a Slack message, or an auto-filed Jira/Linear ticket.

A ready-to-use GitHub Actions workflow (preview vs production, per-page PR comment with the exact tokens that changed, run summary, report artifact, host-auth bypass) is in examples/drift-gate.yml as one full reference. The result-surfacing step (annotations, PR comment) is the only platform-specific part; the extract → compare → branch-on-exit-code core is identical on GitLab CI, Jenkins, and Azure DevOps.

Exit codes

A pipeline can branch on the exit code; "design drifted" and "extraction broke" are distinct:

With --json-only, a failure also prints a machine-readable { "error": { "code", "message" } } to stdout.

Recipes

Quick brand scan

dembrandt dembrandt.com

Compare two sites

dembrandt dembrandt.com --save-output
dembrandt braintree.com --save-output
# Compare output/dembrandt.com and output/braintree.com side by side

Multi-page audit — get a fuller picture across the whole site

dembrandt dembrandt.com --crawl 10 --sitemap --save-output

Spot-check a value — verify a specific token fast

dembrandt dembrandt.com --json-only | grep -i "border-radius"

Export for Tailwind — get spacing and color values into your config

dembrandt dembrandt.com --dtcg --save-output
# Use the .tokens.json with Style Dictionary to generate tailwind.config.js

Export for Tokens Studio / Figma

dembrandt dembrandt.com --dtcg --save-output
# Import the .tokens.json directly into Tokens Studio

Generate DESIGN.md for your AI agent

dembrandt dembrandt.com --design-md
# Point your agent at the output DESIGN.md

Accessibility audit — check contrast on any live URL

dembrandt dembrandt.com --wcag

Regression baseline — snapshot now, catch drift later

dembrandt myapp.com --save-output --dtcg
# Store output as baseline, re-run after deploys and diff

CI / headless environments

dembrandt myapp.com --no-sandbox --save-output

Use Cases

• Design system documentation
• Multi-site design consolidation
• Internal design audits on your own properties
• Learning how design tokens map to real CSS

How It Works

Uses Playwright to render the page, reads computed styles from the DOM, analyzes color usage and confidence, groups similar typography, detects spacing patterns, and returns design tokens.

Extraction Process

1. Browser Launch - Launches browser (Chromium by default, Firefox optional) with stealth configuration
2. Anti-Detection - Injects scripts to bypass bot detection
3. Navigation - Navigates to target URL with retry logic
4. Hydration - Waits for SPAs to fully load (8s initial + 4s stabilization)
5. Content Validation - Verifies page content is substantial (>500 chars)
6. Parallel Extraction - Runs all extractors concurrently for speed
7. Analysis - Analyzes computed styles, DOM structure, and CSS variables
8. Scoring - Assigns confidence scores based on context and usage

Color Confidence

• High: Logo, primary interactive elements
• Medium: Secondary interactive elements, icons, navigation
• Low: Generic UI components (filtered from display)
• Only shows high and medium confidence colors in terminal. Full palette in JSON.

Limitations

• Dark mode requires --dark-mode flag (not automatically detected)
• Hover/focus states extracted from CSS (not fully interactive)
• Canvas/WebGL-rendered sites cannot be analyzed (no DOM to read)
• JavaScript-heavy sites require hydration time (8s initial + 4s stabilization)
• Some dynamically-loaded content may be missed
• Default viewport is 1920x1080 (use --mobile for 390x844 mobile viewport)

Intended Use

Dembrandt reads publicly available CSS and computed styles from website DOMs for documentation, learning, and analysis of design systems you own or have permission to analyze.

Only run Dembrandt against sites whose Terms of Service permit automated access, or against your own properties. Do not use extracted material to reproduce third-party brand identities, logos, or trademarks. Respect robots.txt, rate limits, and copyright.

Dembrandt does not host, redistribute, or claim rights to any third-party brand assets.

Sponsors

The CLI is MIT-licensed and free. Sponsorship funds the enforcement layer: a committed project-level token baseline, --compare and the ingest API for CI/CD drift gates, and the App platform (snapshot history, team drift dashboard, alerts to Slack, Linear, and GitHub).

Contributing

Bugs, weird sites, pull requests. All welcome.

Open an Issue or PR.

@thevangelist

MIT. Do whatever you want with it.