Skip to main content
ai2.design

AGENT-READY API · PRIVATE BETA

Eighteen design APIs, one agent-ready surface.

We did the work behind the Inspiration Gallery — 18-category token extraction, full-page capture, WCAG dual-spec audit, motion library detection, agent prompt kit, atomic registry — and we are exposing every capability as a clean HTTP endpoint. The API surface is in private beta; founding accounts are onboarded in waves and lock today's rate for the entire first year of general availability.

Reserve a tierSee sample response

BUILT FOR

Four audiences, one consistent surface.

  • AI coding agents

    Claude, Cursor, Copilot, Continue.dev — paste an Agent Prompt Kit response into any system prompt and the model writes in the source brand's voice without leaking copy.

  • Design system teams

    Audit a competitor in one call, ship parity recipes the next sprint. Token Extraction plus Multi-format Export bridges Figma libraries to production code with zero manual translation.

  • Headless CMS pipelines

    Webhook on extraction.completed, write tokens straight into a CMS, let editors approve. Theme Generator output drops into Sanity, Contentful, or any token-aware schema.

  • Internal design audit

    Run WCAG + APCA Audit nightly across every brand surface, flag drift the morning of release. Storage API keeps 30-365 day history for compliance evidence.

API CATALOGUE · BETA ROADMAP

Eighteen endpoints across five domains.

The endpoints below are the surface we are shipping during private beta and through general availability (target Q3 2026). Founding accounts get rolling access as endpoints stabilise; quota and pricing lock at signup. Mix and match — each tier ships with pooled quota across the entire surface, no per-API juggling. Namespaced under api.ai2.design/v1.

Extraction

Turn any URL into structured design tokens.

  • Token Extraction
    POST /v1/extract/tokens

    18-category extraction — palette + role + brand-tier flag, typography rules, spacing scale, gradients, shadows, motion timings, z-index ladder, breakpoints, container widths, font-face details, component tree, image profile.

  • Screenshot
    POST /v1/extract/screenshot

    Full-page desktop + mobile + OG image. Selenium scroll-stitch pipeline, gentle mode for animation-heavy sites, sticky-nav neutralization, JPEG q92 with auto post-crop.

  • Brief Generation
    POST /v1/extract/brief

    Curator-grade design brief — verdict, lineage, design decisions, taste notes, signature patterns. Generated from the token extraction pass with optional editorial polish layer.

  • Component Recipe
    POST /v1/extract/recipes

    Per-component Tailwind class plus raw CSS recipes (button-primary, card-surface, link-inline, input-field). Drop into any project as paste-ready snippets.

  • WCAG + APCA Audit
    POST /v1/extract/contrast

    Dual-spec contrast matrix — WCAG 2.1 (AA/AAA times Normal/Large) plus APCA Lc draft WCAG 3. Returns full fg/bg pair grid with sign-aware Lc and verdict label.

  • Motion Library Detection
    POST /v1/extract/motion

    Runtime probe for Framer Motion, GSAP, Lottie, Rive, Lenis, Theatre.js, Spline, Three.js. Window-global detection survives modern bundler obfuscation.

Inspiration

Search the editorial catalogue programmatically.

  • Catalogue
    GET /v1/inspiration

    List all curated items with filters — industry, theme, tag. Paginated, cache-friendly, public-readable on every tier.

  • Detail
    GET /v1/inspiration/{slug}

    Full curator brief plus tokens plus WCAG matrix plus componentRecipes plus agent prompt kit for a single item. Same payload that powers /inspiration/{slug} on this site.

  • Related
    GET /v1/inspiration/{slug}/related

    DNA-based similarity — palette closeness, typography family, motion vocabulary. Returns top-5 with similarity score breakdown.

  • Tag Cloud
    GET /v1/inspiration/tags

    All editorial tags with frequency — useful for autocomplete, faceted browse, or training agent retrieval.

Agent

Paste-ready prompts for AI coding tools.

  • Agent Prompt Kit
    GET /v1/agent/{slug}/prompt-kit

    18-field paste-ready brief — philosophy, mainPrompt, overview, palette/typography/spacing/motion rules, component recipes, fitness checks. Single string ready for Claude / Cursor / Copilot.

  • Design Language Transfer
    POST /v1/agent/transfer

    Source URL plus target project context produces a migration brief. Walks the agent through which decisions to adopt and which to keep, decision by decision.

  • Multi-format Export
    POST /v1/agent/export

    Tokens to four paste-ready formats: Markdown bundle (~2000 word), W3C JSON, CSS variables, Tailwind config object. Single payload, four files.

Launchpad

Atomic components plus theme generators.

  • Atomic Component Registry
    GET /v1/launchpad/atoms

    50+ atomic components (button, input, card, dialog, accordion) with install metadata, dependencies, and copy-paste source. shadcn-compatible registry format.

  • Theme Token Generator
    POST /v1/launchpad/theme

    Palette plus typography seed produces a ready Tailwind v4 plus CSS vars plus W3C JSON bundle. Optional dark-mode mirror generated via OKLCH luminance flip.

  • Starter Generator
    POST /v1/launchpad/starter

    Next.js 16 or Vite scaffolding with selected atoms plus theme tokens pre-wired. Returns zip URL or git push to your repo (OAuth required).

Tooling

Production essentials around every plan.

  • Webhook
    POST /v1/webhooks

    Async job notifications — extraction.completed, brief.ready, screenshot.uploaded. HMAC-signed deliveries with replay protection and exponential backoff.

  • Storage / Versioning
    GET /v1/storage/{job_id}

    Token history retention — 7-day on Maker, 30-day on Builder, 365-day on Studio. Diff any two snapshots to track design language drift over time.

  • Analytics
    GET /v1/analytics

    Per-tier usage dashboard — extractions called, bytes streamed, average response time, top endpoints. CSV export plus Webhook digest option.

PRICING · FOUNDING RATES

Five tiers, one consistent surface.

Quota pools across the entire 18-API surface — there is no per-API metering to juggle. Founding rate locks for the full first year; renewals stay on founding pricing for one additional year.

EARLY ACCESS · LIMITED

Founding builder code — 40% off the first year.

Paste at checkout before your annual term starts. Locks the discounted rate for the entire first year; renewals stay on founding pricing for one additional year. Limited to the first 100 founding accounts across all tiers.

BUILDER40

  • Hobby

    Tinkering with one URL at a time.

    $0forever

    Free · attribution required

    • 50 extractions / month, deferred queue
    • Token Extraction + Screenshot only
    • JSON output (no Markdown / Tailwind config)
    • Public projects only · ai2.design attribution
    Start Hobby

  • Maker

    Indie builders shipping a project at a time.

    $17$29per month

    With BUILDER40 · $209/yr

    Billed annually · $348/yr

    • 1,000 extractions / month, 60 RPM
    • All four export formats (MD / JSON / CSS / Tailwind)
    • Brief Generation + Component Recipes
    • Inspiration API read access (1k req/mo)
    • Email support, 7-day token history
    Start Maker
  • Most popular

    Builder

    Bootstrapped teams running automated audits.

    $59$99per month

    With BUILDER40 · $708/yr

    Billed annually · $1,188/yr

    • 5,000 extractions / month, 120 RPM
    • Full surface incl. Agent + Launchpad APIs
    • Webhooks + Storage (30-day history)
    • Private projects · API key + secret rotation
    • Priority queue, 24h email SLA
    Start Builder

  • Studio

    Agencies and scale-ups running multiple brands.

    $179$299per month

    With BUILDER40 · $2,148/yr

    Billed annually · $3,588/yr

    • 25,000 extractions / month, 240 RPM
    • 5 team seats · white-label exports
    • Storage 365-day history + diff API
    • Linter API + multi-brand audit dashboard
    • Founders Slack channel · 8h SLA
    Start Studio

  • Enterprise

    Volume contracts, SSO, on-prem options.

    Customannual

    Contracted

    • Custom monthly volume + RPM
    • SSO (Okta, Azure AD, Google Workspace)
    • Dedicated infra option · region pinning
    • 99.9% SLA · custom indemnification
    • Quarterly business review · roadmap input
    Talk to founders

SAMPLE RESPONSE

One call, four paste-ready outputs.

Below: an actual Token Extraction request against linear.app, returning the same payload that powers /inspiration/linear on this site. Multi-format Export then converts that single token tree into four distinct artifacts.

bash
curl -X POST https://api.ai2.design/v1/extract/tokens \
  -H "Authorization: Bearer $AI2_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://linear.app",
    "viewport": "desktop",
    "include": ["palette", "typography", "motion", "wcag", "recipes"]
  }'

WHAT YOU GET WITH EVERY CALL

The pipeline behind eight commitments.

  • 18-category extraction

    Palette, typography, spacing, gradients, shadows, motion, z-index, breakpoints, containers, font-faces, component tree, image profile, CSS variables, and more.

  • Brand-tier color flagging

    HSL clustering plus saturation/luminance heuristic identifies which palette entries are brand colors versus neutrals — a signal no other API surfaces.

  • WCAG 2.1 + APCA dual-spec

    Every contrast pair audited against both the legacy ratio formula and APCA Lc draft WCAG 3 — your accessibility report ages well.

  • Motion library runtime detection

    Window-global probe survives webpack/esbuild minification — Framer Motion, GSAP, Rive, Lottie, Lenis, Theatre.js, Spline, Three.js all reliably identified.

  • Curator-grade brief generation

    Verdict, lineage, design decisions, taste notes, signature patterns — the editorial layer no token-only extractor produces.

  • 18-field agent prompt kit

    Philosophy, mainPrompt, overview, palette/typography/motion rules, fitness checks, component recipes — paste-ready for any LLM.

  • Four-format paste-ready export

    Markdown bundle (~2000 word), W3C JSON tokens, CSS variables, Tailwind config — single token tree, four artifacts.

  • Production-grade infra

    Selenium scroll-stitch capture, automatic post-crop, sticky-nav neutralization, gentle mode for animation-heavy sites, 99.9% uptime target on Studio+.

FEATURE MATRIX

Quota across every API

One quota pool spans the entire 18-API surface. Run a Token Extraction on Monday, a Brief Generation on Tuesday, a WCAG Audit on Wednesday — same monthly counter. A-la-carte top-ups available below if you spike.

CapabilityHobbyMakerBuilderStudioEnterprise
Extraction
Token Extraction501k5k25kCustom
Screenshot501k5k25kCustom
Brief Generation—1005002.5kCustom
Component Recipe—5002.5k10kCustom
WCAG + APCA Audit—1k5k25kCustom
Motion Library Detection—5002.5k10kCustom
Inspiration
Catalogue (read)✓✓✓✓✓
Detail (read)1001k5k25kCustom
Related (read)1001k5k25kCustom
Tag Cloud (read)✓✓✓✓✓
Agent
Agent Prompt Kit—1005002.5kCustom
Design Language Transfer——100500Custom
Multi-format Export—1k5k25kCustom
Launchpad
Atomic Registry (read)——✓✓✓
Theme Generator——1001kCustom
Starter Generator——10/mo50/moCustom
Tooling
Webhooks——✓✓✓
Storage / Versioning—7-day30-day365-dayCustom
Analytics dashboard—✓✓✓✓
Team seats1115Custom
Support SLACommunity48h email24h email8h email1h dedicated

A-LA-CARTE TOP-UPS

Spiked this month? Buy quota by the bundle.

Top-ups attach to your active subscription, never expire while the plan is active, and roll over with renewal. Volume above 100k/mo automatically rolls into Enterprise pricing.

  • Extraction bundle

    +1,000 calls

    $9

    Adds 1,000 extractions across Token / Screenshot / Brief / Recipes endpoints. Pooled with your plan's monthly quota.

  • Audit bundle

    +5,000 calls

    $19

    WCAG + APCA Audit + Motion Detection — high-volume nightly compliance runs.

  • Agent bundle

    +500 calls

    $29

    Agent Prompt Kit + Design Language Transfer + Multi-format Export — for high-context migrations.

  • Launchpad bundle

    +50 starters · +500 themes

    $49

    Theme Generator + Starter Generator — agency teams shipping multiple brands per month.

  • Storage bundle

    +365-day retention

    $15 / month

    Extends token history to 365 days regardless of plan. Required for SOC 2 evidence retention windows.

VERSUS THE ALTERNATIVES

Why teams pick AI2 over screenshot APIs and DIY scripting.

CapabilityAI2 DesignRaw screenshot APIsDIY scraping
Design token extraction18 categories + brand-tier flagFonts + OG onlyBuild it yourself
Curator brief generationEditorial verdict + lineage—Manual writeup
WCAG + APCA dual-spec auditBuilt-in contrast matrix—Custom apca-w3 plumbing
Motion library detection8 runtimes (Framer/GSAP/Rive…)—Brittle URL pattern match
Agent prompt kit18-field paste-ready—Glue 5 tools together
Multi-format exportMD + JSON + CSS + TailwindBinary screenshot onlyHand-roll converters
Anti-bot resilienceSelenium + gentle modeStealth on top tier onlyCloudflare blocks you
Effective monthly cost (1k extractions)$17 (Maker founding)$17-$79 capture-onlyDevops time + hosting

API waitlist

Lock the founding rate.

Forty percent off your first twelve months when general availability ships. We open invites in batches by integration depth — tell us what you want to build first.

Interested tier

By submitting you agree to our privacy policy. Response by email — usually within 24-48 working hours.

QUESTIONS

What builders ask before integrating.

  • The full eighteen-endpoint surface is in private beta. Founding accounts that sign up today reserve a quota slot, lock the founding rate for one year of general availability (target Q3 2026), and receive rolling access as each endpoint stabilises. Token Extraction and Screenshot are first; Agent Prompt Kit and Multi-format Export follow shortly after. We do not bill for any tier until at least one endpoint relevant to that tier is live for your account. Full timeline at /legal/api-terms.

EARLY ACCESS

Be among the first to use it.

Free during early access. Ship fast — your agents are waiting.

AI2 Design counts anonymous traffic always — no cookies, no identifiers, no fingerprinting. For optional session insights (PostHog, EU region) — page funnel, performance — turn on opt-in analytics below. Cookie policy.