All work

Case study · 01 of 03

Trcker

Affiliate tracking, owned not rented. A full SaaS competing with Everflow and Tune, built solo in six months.

Site

trcker.io

Year

2024–2026

Role

Solo · everything

Status

Live

Overview

Trcker is a production affiliate tracking platform. Brands run their affiliate programs on it, partners track clicks and conversions through it, and a fraud detection layer scores every postback before it's accepted.

It does what Everflow and Tune do, at a fraction of the price, with an MCP server bolted on so AI agents can pull fraud signals through a public API.

The problem

Affiliate tracking platforms charge enterprise rates. Everflow, Tune, Impact all start at $500–2k a month before you've sent your first click. Startups running their first affiliate program can't justify that.

The alternatives are bad. Open-source trackers lack production polish and fraud features. SaaS link shorteners don't handle postbacks. Building it in-house takes a team a year.

There's a missing tier between hobbyist and enterprise. That's what Trcker fills.

The approach

Strip to the essential pipeline

Click in, conversion out. The rest is decoration. The core tracking pipeline is ~650 lines across 8 files. Everything else is built on that core.

Mandatory postback signatures

Every brand offer has a signing secret. Postbacks without a valid HMAC are rejected at the edge. No exceptions, no quiet failures.

Fraud as a first-class layer

Trcker Radar scores every conversion against patterns (IP velocity, user-agent anomalies, sub-ID stuffing, geo mismatches). Brands see a risk score with the conversion, not after the fact.

MCP server + public API

The fraud detection is exposed as an MCP server so AI agents (Claude, GPT) can query it. Same data also lives behind a REST API. Build partners hook in directly.

Stack

NEXT.JS 16 TAILWIND v4 SHADCN/UI NEON POSTGRES DRIZZLE ORM WORKOS AUTHKIT VERCEL SENTRY POSTHOG PLAYWRIGHT SEMGREP MCP SERVER

Results

6mo

Solo build, idea to production

38+

SEO pages ranking

~650

Lines in core tracking pipeline

Next case

Pulse