Award intelligence · new in v1.7.2 · six tools
Read the winners. Build like them.
Six tools turn Awwwards-grade sites into plans your agent can build: ranked references with reasons, source-backed design DNA, one blueprint onto free Aura components — and audits that say plainly when they're live and when they're static.
The loop · five calls
The loop · five callsReferences in, blueprint out
01Search
ranked refs + try_for cues
aura_award_search { query: "glassy saas" }
02Inspect
design DNA + rule-based ideas
aura_award_inspect { mode: "site_details" }
03Blueprint
brief + tokens + next_calls
aura_design_blueprint { goal: "dark pricing" }
04Audit
capture / bands / compare
aura_visual_audit { mode: "capture" }
05Motion
scans + frame studies
aura_motion_study { mode: "inspect" }The six tools
The six tools29 → 35 tools, zero new deps.
awardaura_award_search
Ranked Awwwards refs across sites, categories, elements or motion — every reason names your word.
awardaura_award_inspect
Source-backed design DNA: palette, tech, elements, awards — with rule-based ideas.
blueprintaura_design_blueprint
Goal + 2–5 refs → brief, tokens_css, section-matched free components and next_calls.
auditaura_visual_audit
Screenshots and band maps when Chrome runs; labeled static-html estimates otherwise.
motionaura_motion_study
Runtime motion scans and bounded frame studies — searchable by library or easing.
awardaura_award_index
Local award corpus: status, rebuild, resume, refresh, winners and watches.
One call · aura_design_blueprint
Four artifacts, in dependency order.
Give it a goal sentence and 2–5 reference slugs. It reads the shared patterns and returns tokens first — so pasting the markup afterward actually holds together.
device, visual direction and constraints are optional inputs — pins, motion and motionless both work.
brief
Goal restated, chosen direction, and the 3–5 shared patterns that actually transfer.
tokens_css
Paste-ready custom properties — paper, ink, accent, radius, type ramp — from the ranked refs.
components
Section-matched free Aura components with ids, so the plan lands on things you can fetch.
next_calls
The follow-ups in order: fetch, install, audit, motion study — no guessing what comes next.
Evidence honesty
Live when it can be. Labeled when it can't.
Live when Chrome runs
aura_visual_audit captures screenshots and band maps. With ffmpeg installed, aura_motion_study scans runtime motion and bounded frame studies.
Labeled when it can't
No Chrome? Estimates are labeled static-html — never dressed up as a live capture. Evidence gaps are named, not rounded over.
Bounded by design
Polite fetching around ~1 request/second, page caps between 1 and 20, and a local corpus under ~/.aura-mcp/awards so re-runs stay cheap.
Chrome and ffmpeg are optional — the award layer degrades to labeled static analysis, never to invented screenshots.
Point it at your favourite winners.
One install command, any MCP client. Then ask for a blueprint and watch the loop run.
npx -y aura-components-mcp@latest