What builders made with Inflectiv: Inside the App Challenge Winners

The App Builder Challenge asked one thing: take a dataset and turn it into something real. Not a slide, not a pitch, a working app running on Inflectiv.

Builders answered. Below are the winning apps, what each one does, and how it uses Inflectiv underneath.

1st place - SiteBrain

Paste your website. Find out why users leave.

SiteBrain is a website intelligence agent built for solo founders. You give it a URL, it reads the page the way a confused first-time visitor would, and it tells you what to fix first. No dashboards, no analytics degree, one report, ranked by what matters.

The insight behind it is sharp. Vibe coding made building an app a weekend job. Understanding why users bounce is still guesswork. SiteBrain closes that gap by turning a website into structured intelligence instead of a hunch.

What it checks

SiteBrain reads ten things a visitor feels but never says out loud: first impression, hero clarity, CTA strength, trust signals, signup friction, pricing confusion, mobile risk, intent mismatch, missing proof, and the quiet conversion blockers that stop people from clicking. Then founders can ask the questions they'd ask an honest friend: why are users leaving, what should I fix first, is my homepage clear, what should I test next?

How it uses Inflectiv

This is the part that made SiteBrain a winner. Every scan becomes a structured intelligence record inside a real Inflectiv dataset, not a text blob, but typed, queryable fields an agent can reason over. The flow runs as a loop: a website scan becomes a structured dataset, a product-intelligence agent reasons over it, recommendations come out, and outcomes write back into the dataset as new intelligence. Each scan makes the next one sharper.

Real scans already write back through Inflectiv's bi-directional API, with the architecture wired to connect any AI agent over the MCP Server. The result is a website that stops being a static page and becomes a living record that keeps learning from real users.

SiteBrain shows off the full Inflectiv pattern in one product: structured datasets, write-back intelligence, and an agent that reasons over data instead of vibes.

Take a look 👉 https://sitebrain-production-6801.up.railway.app/

2nd place - Relay

Drop a chat bubble on any site in one line. It can't make things up.

Most support widgets will happily invent a refund policy on the spot. Relay was built so it can't. You paste a single script tag, and you get a support bubble that quotes only what your team has verified, cited every time, or it hands off to a human. Live in about five minutes, no code.

What makes it different

Three rules run the whole product. Every answer shows its receipts: replies come back with the exact source and a match score, and if there's no source, there's no answer. It only quotes what you approved: sources stay pending until a teammate signs off, so the bubble can't repeat anything your team hasn't put its name on. And it knows when to stay quiet: off-topic or high-stakes questions get a clean handoff to a human instead of a confident hallucination.

Setup is three steps: name your workspace, paste and approve your verified sources, and embed the bubble before the closing body tag on any site.

How it uses Inflectiv

Relay is the grounding gate, and the surface people see. The retrieval brain underneath is Inflectiv. When a teammate approves docs, FAQs, and policies, Inflectiv embeds them into a private dataset. For each question, it returns the exact passages behind the answer, with their source and score attached. Relay's gate then decides: strong source, cite it; nothing solid, refuse and hand off; injection attempt, block it. Never a guess.

Relay went a step further and extended Inflectiv's API to return per-answer source provenance, the list of sources the gate cites on every reply. That's a builder taking the bi-directional API and pushing it to do something new, which is exactly the point of the challenge.

Relay turns Inflectiv's verified-source retrieval into a trust layer any website can install in one line.

Take a look 👉 https://widget-web-production-f2e3.up.railway.app/

3rd place - Treeline

The world's forests, answered from the data.

Treeline is a deforestation intelligence tool. Ask it about tree-cover loss, what's driving it, or whether the world is on track for the 2030 forest pledge, and it answers with the figure, its year, and its source, or it says plainly when the data doesn't cover the question. Every answer leads with a number you can trace, never a guess.

Ask it which countries lost the most primary forest in 2024, and it comes back grounded: 6.7 million hectares of primary forest lost in 2024, up 80% versus 2023, with fire ranking as the top driver for the first time.

What makes it work

Treeline runs on verified records only. It pulls from Global Forest Watch and FAO data, and every response carries the figure, the year, and the source behind it. When a question falls outside what the data holds, it says so rather than filling the gap. It also draws a clear line on scope, it reads forest data and answers forest questions, and it stays out of investment and carbon-credit advice entirely.

How it uses Inflectiv

Treeline is a clean example of retrieval done right on Inflectiv. Global Forest Watch and FAO records live as a structured dataset, every figure attributed to its year and source. When someone asks a question, Inflectiv returns the exact records behind the answer, and Treeline surfaces the number with its provenance attached. The honesty of the product comes straight from the architecture: because the data is structured and attributed, Treeline can tell the difference between what it knows and what it doesn't, and answer accordingly.

Treeline turns a pile of public forest data into a source you can question directly and trust, because every answer shows where it came from.

Take a look 👉 https://treeline-production-6200.up.railway.app/

Three winners, three very different products, one shared foundation.

SiteBrain writes intelligence back into a dataset that keeps learning.

Relay reads verified intelligence out with full provenance on every answer.

Treeline grounds every response in attributed records and refuses to guess.

All three prove the same thing: a dataset on Inflectiv is not a file sitting on a shelf. It's queryable, ownable intelligence that real apps can build on.

The App Builder Challenge is over. The Startup Program stays open, with free credits, marketplace datasets, and a showcase for whatever you build next.

Build on it → https://www.inflectiv.ai/campaigns