Plain guide

What is MCP for business?

MCP, the Model Context Protocol, is an open standard that lets AI clients like Claude, Cursor, and ChatGPT connect to real tools. Riley exposes your business as an MCP server, so an AI can do the work while every external write still stops for your approval.

Start free

MCP, the Model Context Protocol, is an open standard that lets AI clients like Claude, Cursor, and ChatGPT connect to real tools instead of just chatting. For a business, it means an AI you already use can reach your actual work through a connection you control.

Riley exposes your business as an MCP server. A connected AI can prepare work through Riley's tools, and any action that writes outside your workspace still stops for your approval, then lands on a signed receipt you can share.

Who this is for

For teams that already live in an AI client

MCP matters when the AI you like is not where your business tools are.

Agencies

Agencies with an AI workflow

Drive Riley from the AI client your team already works in, and keep the same approval rules across every client account.

Local business

Local businesses

You do not have to learn a new interface. Ask Riley for work from a tool you already use, or just use the dashboard.

Ecommerce and SaaS

Ecommerce and SaaS teams

Connect the AI clients your team standardized on, and bring your own tools in through Riley's external tools support.

Consultants and builders

Consultants and builders

Point a technical client at the MCP endpoint and build your own flows on top of Riley's tools, with keys you control.

The problem

Great AI, locked out of your real work

The AI clients people love are chat boxes. They can reason and write, but they cannot open your reviews, edit a listing, or send a follow-up, because they are not connected to any of it. So the person becomes the connector, copying answers back and forth by hand.

Before MCP, closing that gap meant a custom integration for every tool and every client, which few businesses have time to build. MCP is the shared standard that removes the custom work. Connect an MCP-speaking client to an MCP server once, and the AI can use those tools directly.

How Riley helps

Riley is the MCP server for your business

Connect the AI client you like, and it works through the same loop and the same rules as the Riley dashboard.

  • Connect your tools to Riley, and Riley to your AI client over MCP.
  • Riley runs the work you approve, no matter which client asked.
  • It verifies each change on the live source.
  • It signs a proof receipt, whether the request came from Slack or an MCP client.
Where it connects

AI clients you can connect over MCP

Honest statuses only. Verified clients are documented end to end; others connect with guided setup we have not individually verified.

Claude Desktop

Verified

Add Riley as an MCP server and work from Claude Desktop.

Claude Code

Verified

Add Riley with one command and use it from the terminal.

Cursor

Verified

Add Riley as an MCP server in Cursor settings.

ChatGPT

Where supported

Works where ChatGPT supports custom MCP connectors on your plan.

Other MCP clients

Guided setup

Many MCP-compatible clients connect with guided setup, not individually verified.

Your own tools

Live

Bring your own MCP server or import an OpenAPI spec, and Riley uses your tools inside the same rules.

A concrete example

Driving Riley from Claude

From a chat window to real, approved work

  1. You mint a Riley key in Settings, where it is shown once.
  2. You add Riley as an MCP server in Claude Desktop and paste the endpoint and key.
  3. You ask Claude to have Riley draft a reply to a new review.
  4. Because posting is an external write, it queues for your approval just like any other Riley work.
  5. You approve, Riley posts and verifies it, and it lands on your signed receipt.

The key is sent as an x-api-key header, it is per workspace, and you can revoke it in Settings anytime. Full setup steps live on the developers page.

The proof it creates

Same rules, same receipts, any client

Approval history

Work from any connected client queues for approval, and each decision is recorded like any other Riley action.

Signed receipts

A request from an MCP client shows up on the same signed weekly receipt as everything else.

Per-key access

Each client connects with its own per-workspace key, so you can see which one connected and revoke it anytime.

Common questions

MCP for business FAQ

Which AI clients can connect to Riley over MCP?

Claude Desktop, Claude Code, and Cursor are verified today. ChatGPT works where it supports custom MCP connectors on your plan. Many other MCP-compatible clients can connect with guided setup, though Riley has not individually verified each one. You connect any of them with a per-workspace key you mint and can revoke.

Is it safe to connect an AI client to my business over MCP?

The connection uses a per-workspace key that is shown once and can be revoked at any time in Settings. Whatever client you connect gets only Riley's scoped tools, and any action that writes outside your workspace still queues for your approval. So a connected AI can prepare work, but it cannot change a real account without your yes.

Do I need to be technical to use MCP with Riley?

Not for the everyday product. Most people just use Riley in the dashboard or in Slack. MCP is for when you want to drive Riley from another AI client you already work in. Setup is a few steps: mint a key, add Riley as an MCP server, and point the client at the endpoint.

Can I connect my own tools too?

Yes. You can bring your own MCP server or import an OpenAPI spec from the external tools area of your dashboard. Riley then uses those tools inside the same approval rules, and writes through them queue for your approval just like everything else.

Connect your AI. Keep the guardrails.

Drive Riley from the AI client you already use, with your approval on every write. Start free, no card to begin.