AI: Extract from URL

What it does Given a website URL, extracts a full first draft brand kit: name, description, mission, brand story, taglines, personality traits, tone of voice, visual style, products, and up to 5 detected target audiences. How it works 1.

Firecrawl scrapes the site to markdown. 2. A single AI call parses the markdown into a structured brand kit payload. 3. A follow up call proposes candidate audiences from the same scrape. 4.

You approve/deny each detected audience; approved ones queue background enrichment. Model via Lovable AI Gateway. Fields written Across , , , , , , and (via detect) . Cost 1 credit — URL extraction (fills brand kit core + expression + products).

+1 credit per accepted target audience that runs enrichment. Declined proposals are free. Where to trigger it Create Brand Kit → Extract from URL tab (default tab for new brand kits). MCP equivalent Not exposed today.

MCP clients construct brand kits with the section scoped write tools. Failure modes Site blocks scrapers → Firecrawl returns empty markdown; the AI has no context and returns thin output.

JavaScript only content → markdown is sparse; retry with a static/blog URL if available. Rate limited → surfaces as a 429 in the AI Workflows dashboard.