Connect an AI assistant (MCP)

Connect Claude or ChatGPT to your account over the Model Context Protocol: one address, no key to copy, what you are asked to approve, and how to disconnect one.

MarketplaceHub speaks the Model Context Protocol (MCP), so an AI assistant can read your account directly instead of you copying answers into it. Ask Claude or ChatGPT how many units of a SKU are left, which marketplace connection is failing, or whether last night's price push finished, and it reads the answer from your own data.

It is the same data the API serves, through the same permissions and the same request allowance. Nothing new is exposed.

Connect one

There is one address:

https://api.marketplacehub.com/mcp

Add it as a custom connector in your assistant — in Claude, Settings → Connectors → Add custom connector; in other clients, look for "MCP server" or "custom connector". Paste the address and the assistant does the rest.

You do not create an API credential for this. Unlike the API pages, the assistant registers itself and then sends you to a normal MarketplaceHub sign-in page. There is no secret to copy and nothing to paste back.

What you are asked to approve

After you sign in, MarketplaceHub shows a consent screen listing exactly what the assistant asked for — for example Read products and listings. Read it. Approving is what creates the access; nothing happens before you do.

• An assistant gets only the permissions it asked for, and can never be granted more than your account offers.
• Access is per assistant. Connecting Claude does not connect anything else.
• One item is not a permission over your data: Stay connected. It lets the assistant keep working without sending you back to a sign-in page every hour. It reads nothing extra, and disconnecting ends it too.

What it can do

Eleven tools. Nine only read — your products and their listings, marketplace categories and the attributes they need, which marketplaces are connected and how healthy each one is, whether a background job finished, and your orders.

Two can change things, and an assistant gets those only if you approved the matching write permission:

• Set price and stock for SKUs you name, and push the change out to your connected marketplaces.
• Confirm a shipment of order items, with carrier and tracking.

An assistant you approved for reading is never even offered them.

What it deliberately cannot do

• Publish a new listing. Publishing writes to a marketplace against rules that have to be exactly right, which is a poor first thing to hand an assistant.
• Refund a buyer. Money moves, and it does not move back on its own. Refunds stay with a person, on the Orders page.
• Ship an order the marketplace fulfils itself. A confirmation on an Amazon FBA order is refused, because Amazon is already doing it.
• See your buyers by accident. A buyer's name and address are left out of every order answer unless the assistant asks for them on purpose.

If something else in your account owns price, stock or orders, a change made through an assistant is still applied — and the answer names the owner, so the assistant can tell you rather than assume. See the Data Flow page.

Repeats cannot double up

An assistant that never sees the answer to a change may send it again. Both write tools are protected against that: a repeat of the same call returns the first answer instead of doing the work twice, so a lost reply cannot confirm one shipment twice or apply the same price change again. You do not have to switch anything on.

Stop an assistant

Open API → Assistants. Every assistant you have approved is listed with what it was allowed to do and who approved it. Disconnect removes it.

It then has to ask for your approval again before it can read anything. A token it already holds keeps working until it expires, up to an hour — so if you think a token has leaked, contact support as well.

The list belongs to the account rather than to one person, so a colleague can disconnect an assistant you connected, and you can disconnect theirs. Removing the connector inside the assistant stops it from their end; disconnecting here stops it from ours.

If it will not connect

• Nothing happens when you paste the address. Check your client supports remote MCP servers with sign-in. Desktop and browser assistants generally do; older command-line tools may not.
• You sign in but the assistant reports an error. The usual cause is the sign-in window being closed before it finishes. Try again and let it complete.
• A tool says a permission is missing. The assistant asked for less than your question needs. Reconnect it and approve the wider set on the consent screen.

Where to go next

• Using the API — credentials, quotas, logs and webhooks.
• MCP reference — every tool, and the permission each one costs.