Our story
We decode brandsso you don't have to.
BrandRNA exists because extracting brand identity from a website shouldn't take a designer, a front-end engineer, and three meetings. One URL, one API call, one complete brand pack.
Specimens in rotation
Real brands. Real packs.
A live sample of brands already decoded by the API — logos, palettes, and identity, all structured.
Three steps
What we do.
From URL to structured brand intelligence - no manual work, no guesswork.
01
Extract
Drop in any public URL. Our Chromium engine renders the page like a real browser, capturing every CSS variable, font declaration, logo asset, and screenshot - the full visual fingerprint.
02
Analyze
A multi-agent LLM pipeline runs in parallel - colour, typography, logo, style, and validator agents cross-check every finding. The result is a structured brand pack, not a guess.
03
Amplify
Get a complete structured JSON response, a REST API, an MCP server for AI agents, and downloadable asset bundles. Plug BrandRNA into your landing page personalisation pipeline, CMS, or design toolchain.
The pipeline
How it works.
From URL to validated brand pack in under a minute. The high-level flow - no PhD required.
01
Submit a URL
Send any public website URL to our API. Your application, CI pipeline, or AI agent makes a single HTTP call - no SDK, no setup.
02
We render the page
Headless Chromium loads the full page including webfonts and dynamic content. We capture the computed CSS, screenshots, and all `<link>` / `<meta>` / manifest data.
03
Multi-agent extraction
Five specialised LLM agents run concurrently: colour, typography, logo, style, and a validator agent that cross-checks every result for accuracy and consistency.
04
Validated brand pack
A structured JSON response with colours (with roles), fonts (with weights + sources), logos (dark / light / transparent variants), brand voice, similar brands, and industry norms.
Production-grade
The technology.
Built for production from day one - not a prototype, not a demo.
Built on open standards
FastAPI serves the pipeline. Playwright drives headless Chromium. OpenRouter routes to multiple frontier LLMs so you get the best model per agent, not a single-model compromise. Results are cached for 24 hours so repeat lookups are instant and nearly free.
Production-grade from day one
Cloud Run scales to zero when idle and spikes to meet demand in seconds. Firestore stores structured data. Stripe handles metered billing with per-call microtransactions. No minimums, no subscriptions - pay only for what you extract.
What we believe
Our values.
Speed
Seconds, not days. Get a complete brand pack before your CI pipeline finishes. Cached lookups return in under 100ms.
Accuracy
Validated, not guessed. Every finding is cross-checked by a second LLM. CSS parsing and k-means clustering provide deterministic anchors.
Simplicity
One endpoint, endless possibilities. No SDK to learn, no schema to study. curl it, pipe it to jq, ship it to production.