DEV Community

pipeyard
pipeyard

Posted on

Why I built an MCP marketplace for industry verticals

Every time I tried to find an MCP connector for the software
my clients actually use, I hit the same wall.

There are plenty of lists of MCP servers out there. GitHub,
Notion, Slack, Linear — all well documented and easy to find.
But what if you're building an AI agent for a construction
company? A hospital? A logistics firm?

Try finding a working, documented MCP connector for Procore,
Epic EHR, QuickBooks, or ShipBob. Good luck.

That's the problem I kept running into — and it's why I built
Pipeyard.

The problem with generic MCP lists:

Most MCP server catalogs are built for developers working on
general productivity tools. They cover the obvious ones well.
But real businesses run on industry-specific software — and
those connectors are scattered, undocumented, or just don't
exist yet.

If you're building agents for real clients in construction,
finance, healthcare, or logistics, you're basically starting
from scratch every time. You have to read through API docs,
figure out auth flows, guess at response formats, and build
a thin MCP wrapper yourself.

That's weeks of work before you've even started on the actual
agent logic.

What Pipeyard solves:

Pipeyard is a curated MCP connector marketplace focused on
industry verticals. Instead of generic tools, every connector
on Pipeyard is built for the software that businesses in
specific industries actually use.

Right now it covers 4 verticals:

  • Construction — Procore, Autodesk Build, Buildertrend
  • Finance — QuickBooks, Xero, FreshBooks, NetSuite
  • Healthcare — Epic EHR, Cerner, Athenahealth
  • Logistics — ShipBob, ShipStation, Samsara, FedEx

30+ connectors total, with more being added based on
community demand.

How it works:

Every connector on Pipeyard comes with everything you
actually need to use it:

Full credential setup guide — step by step instructions
for getting your API keys or OAuth credentials, specific to
that tool.

Real curl examples — copy-paste ready commands you can
run immediately to test the integration before writing a
single line of agent code.

Example JSON responses — so you know exactly what data
structure your agent will receive back. No surprises.

Sandbox testing — a built-in test panel on every
connector page where you can run a prompt and simulate what
the connector does.

Security score and rate limit info — the practical
details that matter when you're building production agents.

The first real hosted MCP server — for Procore — is live on
GitHub with full TypeScript source code, covering projects,
RFIs, submittals, documents, and budget tools.

What's next:

Pipeyard is currently in beta and completely free to use.

The roadmap is community-driven. There's a connector request
page where you can submit the software you need and vote on
other requests — that's how I'm deciding what to build next.

If you're building AI agents for real industries and you've
hit the same wall I did, I'd genuinely love your feedback.

Try it here: https://pipeyard-qx5z.vercel.app

Top comments (0)