What RuleSell is and why we built it

April 28, 2026/RuleSell Team

We built a verified marketplace for AI dev assets because star ratings are broken and download counts measure marketing, not quality. Here's how RuleSell works and why.

If you use Claude Code, Cursor, Windsurf, Cline, Codex, or any of the other AI coding tools, you've run into this problem: there's a lot of community content — rules, skills, MCP servers, agents, workflows — and most of it is 60% there. The author tried it once on their project, it worked, they published it, and now their 3-month-old cursorrules file is breaking your cursorrules file.

There's no good way to tell ahead of time which ones will work. Star ratings are mostly "the author's friends liked it". Download counts don't mean usefulness. Community-voted rankings get gamed. Quality slops.

We spent too many nights on this. So we built RuleSell.

Two things we think are different

Quality is measured, not voted

Every listing on RuleSell has a Quality Score — a number between 0 and 100. It's composed of six signals, all measured automatically:

Trigger reliability — does the description actually fire on real user

phrases? We test against a curated corpus.

Token efficiency — how many tokens does it cost to load + use? A great

skill that costs 5000 tokens per activation is worse than a good one that costs 500.

Schema cleanliness — does the frontmatter, manifest, or MCP capability

definition actually validate?

Install success rate — does it work first try on a fresh machine?
Freshness — when was it last updated? AI tools move fast and stale

configs rot.

Security scan — does it touch credentials, make unexpected network calls,

or fail a static analysis pass?

Star ratings don't control the sort order. Download counts don't. Only these six things do. You can read the full scoring model at our Trust & Quality page.

Every paid item passes a verification gate

To publish a paid item on RuleSell, a creator must:

Verify email + 18+
Publish at least one free item and pass a basic scan
Accumulate 50 verified installs on their free work
Complete Trader KYC (DSA Article 30 identity verification — required for EU

sellers anyway)

Connect a Stripe Connect account for payouts

This is slow. On purpose. It's how we keep the drive-by spam that hit some other creator marketplaces out of ours. Real creators clear the bar in 1-3 weeks. Drive-by scammers don't bother.

Who this is for

Developers looking for high-quality AI configs. You want to know that when

you install something, it'll actually work, not break your week.

Creators of AI dev tools. You want to get paid for the thing you'd build

anyway. 85% rev share, monthly payouts, real payouts (not "we'll pay you when we hit a milestone").

Maintainers of open-source dev tools. You want to claim your content

officially, show maintainer verified badges, and have a first-party place to point people to.

The business model

Creators keep 85%. 90% for the first six months (a launch promo).
Refunds are 14-day no-questions for items not yet installed.
Chargebacks eat seller revenue (plus a $5 admin fee) — we protect the

buyer first.

Payouts are monthly, $50 minimum, via Stripe Connect.
We use Merchant of Record (LemonSqueezy at launch, migrating to Stripe +

Paddle over time) to handle EU VAT, UK VAT, and US sales tax. You don't have to deal with that.

What's on the platform at launch

10 Official listings hand-built by our team, covering the main categories

(plugin starter, MCP server template, hook presets, Cursor rules for React 19, Prisma workflow, AGENTS.md templates, Stripe Checkout rules, shadcn conventions, TypeScript debug skill)

~50 community listings from Founding Creators we've recruited personally.

Every Founding Creator has a badge on their profile.

Support for Claude Code, Cursor, Windsurf, Cline, Codex, Gemini CLI, ChatGPT,

Copilot, n8n. One catalog, multi-environment.

What's not on the platform yet

No team/organization features beyond basic multi-seat
Paid items haven't launched yet (we're running a free-only soft launch for the

first 2 weeks to bootstrap the trust graph)

The full quality scoring pipeline uses v1 fixture data — v2 will use live

telemetry from real installs

Full subscription item support is coming in month 2

What we want from you

If you try RuleSell and have feedback — good, bad, anything — send it to us. The first 30 days of feedback will shape what we build next. We read every comment, every issue, every tweet.

If you publish dev tool configs and want early access as a Founding Creator, DM us. You'll skip the install gate, get a "Founding Creator" badge, and we'll co-promote your work.

If you install something that breaks, open an issue. We'll make it right.

Who built this

Two people (Nalba and Baha) and a lot of Claude Code. We're based in Turkey, shipping from Istanbul. RuleSell is our first serious marketplace project — we've built and sold software before, but the combination of trust infrastructure, creator payouts, multi-tool catalog, and regulated compliance is the biggest challenge we've taken on.

The goal is simple: make AI dev tools feel less like a wild west and more like a place where quality work gets paid and slop gets filtered. One verified listing at a time.

— RuleSell team

Read more about how we think about quality: How we measure quality, Why prompt engineering matters more than model selection, and The real state of MCP servers in 2026.