← Projects

PourPilot for xBloom

June 2026
Next.jsVercel AI GatewayNeonopencodeClaude

Snap a coffee bag, get an AI pour-over recipe, send it to your xBloom

Snap any bag. Get the perfect recipe.

  • Snap & read

    A vision model reads the bag — origin, process, roast, tasting notes.

  • Smart lookup

    Sparse label? It searches the web or reads the roaster's product page.

  • Hot, iced & cold brew

    Pour-over, iced, or cold brew — it designs the right recipe for each.

  • Tune by taste

    Too bitter or sour? It adjusts grind, temp, and ratio like a barista.

  • One tap to xBloom

    The finished recipe syncs straight to your machine, ready to brew.

Your AI co-pilot for pour-over

My xBloom Studio brews to a recipe, but dialling one in by hand means typing dose, ratio, grind, temperature, and every pour into the app — so most of us just guess.

PourPilot is the shortcut: snap the bag, and AI reads the label (or looks the coffee up online) and designs a pour-over grounded in real extraction theory. Tweak it, then send it straight to your xBloom, ready to brew.

xBloom has no public API, so it talks to the same private endpoints the app uses (RSA-encrypted, building on denull0/xbloom-agent). AI runs through the Vercel AI Gateway, recipes live in Neon, and it's a self-host template. Built end to end with opencode.


github.com/hgstrm/pourpilot, photo to recipe to your machine.