If you have used ChatGPT or Claude for product research, you have probably hit the same wall. The assistant writes well, but it can't see your market. It doesn't know today's Amazon prices, how a product's sales rank moved last quarter, or what your supplier charges. The Model Context Protocol, or MCP, is the standard that closes that gap.
This guide explains what MCP is in plain English, what it changes for ecommerce sellers, which apps support it and how to connect your first data source.
MCP in plain English
MCP is an open standard that lets AI assistants connect to external tools and data. Anthropic introduced it in late 2024 and published it openly, so any company can build on it. Many AI apps now support it.
A simple way to picture it: before MCP, each AI app needed its own custom integration for every service. With MCP, a service builds one connection, called an MCP server, and any app that supports MCP can use it. It works like a standard plug: the same plug fits many sockets.
Once connected, your assistant can call the service's tools during a conversation. You ask a question in normal language. The assistant decides which tool to use, sends the request, reads the result and explains it to you.
The four parts, explained
| Term | What it means | Seller example |
|---|---|---|
| MCP client | The AI app you talk to | Claude, ChatGPT, Claude Code, Cursor |
| MCP server | The service that offers tools and data | A product research service |
| Tool | One action the server can perform | Research products, calculate profit |
| Sign-in | How the server knows who you are and what the assistant may do | Logging in with your account and approving access |
You don't need to build anything. As a seller, you add a server to an app you already use, sign in and start asking questions.
Why MCP matters for sellers
Real numbers instead of plausible ones
A language model on its own predicts likely text. Ask it for the monthly search volume of a keyword and it may produce a precise-looking figure with no source. With an MCP connection to a data service, the assistant can fetch the figure and tell you where it came from. You still check it, but now there is something to check. We cover this problem in more detail in product research with ChatGPT or Claude.
Fewer tabs and less copying
Research often means moving between a sales-rank tool, a trends tool, a spreadsheet and a fee calculator. With MCP, you can ask one question and the assistant calls several tools in turn, then puts the results side by side.
Your data stays where it lives
The assistant reads from the service when it needs to, so you don't paste exports into every chat. Saved products, watchlists and alerts stay in the service, and you can pick them up from any connected app.
What a conversation looks like
Here is an illustrative example of a request with a product research server connected:
Find kitchen storage products for Amazon US under $30 with steady demand and low competition.
Behind the scenes, the assistant:
- Picks the research tool and fills in the details from your request: category, market and price limit.
- Tells you the tool spends credits and asks you to confirm.
- Starts the research, which runs in the background, and checks back until it is finished.
- Summarises the ranked results, with a source for each number.
You can then follow up with "Validate the second one with a $4.50 unit cost" or "What's my margin at $24.99 on Shopify?" The assistant keeps using the tools as the conversation goes on.
Which apps support MCP
Support keeps growing. The main ones sellers use:
- Claude (web, desktop and mobile). Add a custom connector in the settings. On Team and Enterprise plans, an owner adds it for the organisation first.
- ChatGPT. Custom MCP servers are added through developer mode and connectors. Which plans can add them, and what they may do, is set by OpenAI and depends on your plan.
- Claude Code. Anthropic's agent for the terminal. One command adds a server.
- Cursor, Codex, VS Code and Gemini CLI. Each has its own settings file or command.
If you mostly work in a chat window, start with Claude or ChatGPT. If you like keeping research notes and spreadsheets in a folder, a command-line agent works well. See our guide to product research with Claude Code.
Is MCP safe?
An MCP server can only do what its tools allow, and only after you connect it. Good practice:
- Connect servers you trust. A server sees the requests your assistant sends to it.
- Check what you approve. Many servers let you choose permissions at sign-in. Ecomtoro, for example, lets you leave out credit spending for a read-only connection.
- Keep confirmations on. Most clients ask before running a tool. Read the request before you approve it.
- Disconnect when you're done. You can remove a connector in the app and revoke access in the service's settings.
What MCP does not do
MCP is a connection, not a guarantee of good answers. The assistant can still misread a result or draw a weak conclusion. The data is only as good as the server behind it. And the decision to order stock is still yours.
Treat MCP as a way to bring evidence into the conversation, then check the evidence. Our validation checklist explains what counts as evidence at each step.
How to connect Ecomtoro over MCP
Ecomtoro runs an MCP server at https://ecomtoro.com/api/mcp. MCP access is included on every plan, including Free, with the same credits and limits as the app.
In Claude: open Customize, then Connectors, choose Add custom connector and paste the URL. Sign in to Ecomtoro when asked and choose what Claude may do.
In Claude Code: run this in a terminal, then type /mcp inside Claude Code, pick ecomtoro and choose Authenticate.
claude mcp add --transport http --scope user ecomtoro https://ecomtoro.com/api/mcp
Setup for ChatGPT, Cursor, Codex, VS Code and Gemini CLI is on the MCP setup page.
Once connected, your assistant can use these tools:
| Tool | What it does | Cost |
|---|---|---|
| research_products | Finds and ranks product opportunities in the US, UK or Germany | Credits, asks first |
| validate_product | Builds a dossier with demand, trend, competition, margin and risk scores | Credits, asks first |
| get_run | Checks the progress and result of a research or validation run | Free |
| calculate_profit | Profit, margin, ROI and break-even ROAS, including VAT, duty and fees | Free |
| save_product, list_products, get_product_dossier, compare_products | Save, view and compare products | Free |
| add_to_watchlist, schedule_research, list_watchlist, get_alerts | Track products and niches and read alerts | Free |
Every number in a research or validation result comes with a source, so you can check it. For a wider view of what else AI can take off your plate, read AI for ecommerce: what sellers can automate today.
Frequently asked questions
What does MCP stand for?
MCP stands for Model Context Protocol. It is an open standard, introduced by Anthropic in late 2024, that lets AI assistants connect to external tools and data.
Do I need to code to use MCP?
No. In Claude and ChatGPT you add a server URL in the settings and sign in. Command-line tools like Claude Code need one command, which you can copy from the service's setup page.
Does ChatGPT support MCP?
Yes, through developer mode and connectors. Which plans can add custom MCP servers, and what those servers may do, is set by OpenAI and varies by plan, so check your own settings.
Is MCP the same as an API?
Not quite. An API is how software talks to a particular service. MCP is a shared standard for AI assistants to use services' tools, so one MCP server works across many AI apps without a separate integration for each.
