Corey Corey
How it worksWhat Corey buildsCase studiesYour dataDocsStart with Corey
How it works

You steer. Corey does the work.

Describe what you want in plain words. Corey plans it, does it across the tools you already use, and comes back for your approval before anything ships. Here is the whole thing, end to end: the connection, the loop, and where you stay in control.

The model

Engine plus overlay.

Two parts. Corey ships with neither knowledge of nor opinions about your business. You bring the second part.

The engine

Corey. The reasoning, planning and execution core - generic and identical for every buyer. It drives tools but knows nothing about your work.

Your overlay

Your facts, voice, products, tools and guardrails - the brief that turns a capable engine into your operator, built once and refined as you go.

Your operating system

Engine plus overlay, running. A system that does the operational work the way you would, and reports back before anything ships.

generic
Engine
+
yours
Overlay
=
running
Your OS
The connection

Corey runs as an MCP server.

You connect Corey to your stack from inside Claude - Desktop or Code. No platform to migrate to, no data to hand over - Corey reaches into the tools you already use.

01

Add the server

One line in Claude Code, or a custom connector in Claude Desktop, registers Corey as an MCP server.

02

Sign in, once

A Corey sign-in page appears. Signing in once is your identity and creates your account - no card, no second login.

03

Talk to it

Get started, then describe the work in plain words. The guided intake starts your first 28 days on us - all in conversation.

claude code
# register Corey as an MCP server
claude mcp add --transport http corey https://mcp.getcorey.ai/mcp --scope user

# then just ask
corey, triage today's issues and draft replies for the questions
The loop

Plan, execute, review. You steer.

Corey never goes straight to doing. Every job runs the same loop, and you hold the approvals that matter.

It writes the plan first

Corey breaks the request into steps, names the tools it will touch and shows you the plan before acting. If the plan is wrong, you correct it in a sentence.

It does the work

Corey runs each step over MCP, calling your connected tools. Read-only steps run freely; anything that writes, sends or spends waits behind an approval you set.

You approve, it reports

Corey hands back drafts, flags and a summary of what it did. You approve, edit or reject. It records the outcome and folds your feedback into the overlay.

i
Corey executes with a human steering. It plans and does multi-step work end to end, but it does not ship the things that matter without your nod. That is the point.
Questions

How it works, answered.

Corey runs as an MCP server. You add it from inside Claude (Desktop or Code) with one line, sign in once, then talk to it in plain language. It reaches into the tools and accounts you already use - there is no platform to migrate to and no data to hand over.
The engine is Corey: the reasoning, planning and execution core, identical for every buyer. The overlay is yours: your facts, voice, products, tools and guardrails. Engine plus overlay, running, is your operating system - a system that does the work the way you would.
No. Corey plans every job and shows you the plan before acting. Read-only steps run freely, but anything that writes, sends, spends or is irreversible waits behind an approval you set. You stay the CEO; Corey is the operator.
No. After the one-line connect and a single sign-in, everything happens in conversation. You describe the work in plain words and approve what ships - no code, no infrastructure to manage.

This is not a demo. Corey built and runs this site, runs Presso Network Ltd day to day, and is building Safebites and Emilista alongside it. One other founder runs their business on Corey today. What you are reading is its work - see the case studies.

See the loop on your own work.

Connect Corey, hand it one task, and watch the plan-execute-review loop run for real.