Vibe coding · MCP servers

Claude MCP: connect Claude to your tools and data

Claude MCP is how you connect Claude to the tools and data you already use. MCP stands for Model Context Protocol, an open standard. You add an MCP server for a tool, and Claude can then read from it and act in it. Think of it as a USB-C port for AI.

What is an MCP server?

An MCP server is a small program that sits between Claude and one tool or data source. It tells Claude what it can do there, then does it when Claude asks. Anthropic's docs put it this way: MCP servers give Claude Code access to your tools, databases and APIs.

One server, one job

A server usually covers one thing: your notes app, your database, your design files or your task board.

An open standard

MCP is open source. A server written once can work with any AI app that speaks MCP, not only Claude.

Remote or local

Some servers run on the web and you connect with a link. Others run as a small program on your own computer.

You stay in control

You choose which servers to add, and you can remove one at any time.

What you can do with Claude MCP

Work from real data

Let Claude read your database or your docs, so it stops guessing what is in them.

Act in your tools

Claude can create the task, update the record or post the note, instead of telling you to do it.

Build faster

While you build, Claude can check the real database, the design file or the error log on its own.

How to add an MCP server to Claude Code

These steps follow Anthropic's Claude Code docs, as at 10 November 2026. The exact commands are in the table below.

  1. 01

    Find the server

    Get the server's web address or install command from the tool's own site, or from Anthropic's connector directory.

  2. 02

    Add it

    Run the claude mcp add command with a name and the address. Web servers use the HTTP transport.

  3. 03

    Sign in if asked

    Many servers ask you to log in to the tool. Type /mcp inside Claude Code to see each server and its status.

  4. 04

    Ask for something real

    Give Claude a task that needs the tool. Watch what it does the first few times.

Where a server is saved

Claude Code saves each server in one of three scopes. The scope decides who gets it. Local is the default.

Local

Only you, only in the project where you added it. Good for trying a server out.

Project

Saved in a .mcp.json file in the project, so your whole team gets the same servers.

User

Only you, in every project on your machine. Good for tools you use everywhere.

Stay safe with MCP servers

An MCP server can read and change real things. Treat adding one like giving a new staff member a login.

Only add servers you trust

Anthropic's docs say to verify you trust each server before you connect it.

Watch for hidden instructions

A server that fetches outside content can carry text written to trick the AI. Anthropic calls this prompt injection risk.

Give the least access

If Claude only needs to read, connect a read-only account. Keep write access for when you need it.

Review project servers

When you open someone else's project, check which servers it wants to add before you approve them.

★ Checked 10 November 2026

Claude MCP commands for Claude Code

The commands you will use most. Swap in your own server name and address.

Claude Code MCP commands and what each one does
What you wantCommand
Add a web serverclaude mcp add --transport http <name> <url>
Add a local serverclaude mcp add [options] <name> -- <command> [args...]
Share it with your projectAdd --scope project to the command
Use it in all your projectsAdd --scope user to the command
List your serversclaude mcp list
Check one serverclaude mcp get <name>
Remove a serverclaude mcp remove <name>
See status inside Claude Code/mcp

From Anthropic's Claude Code docs, read on 10 November 2026. Anthropic recommends the HTTP transport for remote servers and says the older SSE transport is deprecated.

Questions, answered

What is Claude MCP?

Claude MCP is Claude's support for the Model Context Protocol. MCP is an open standard for connecting AI apps to outside systems. With it, Claude can use your tools, files and data through MCP servers.

What is an MCP server in simple terms?

An MCP server is a small program that links an AI app to one tool or data source. It lists what the AI can do there and carries out the request. One server for your database, one for your task board, and so on.

How do I add an MCP server to Claude Code?

Run claude mcp add with a name and the server's address. For a web server the form is: claude mcp add --transport http, then the name, then the URL. Type /mcp inside Claude Code to check it is connected.

Can I use MCP in the Claude apps, not just Claude Code?

Yes. In the Claude apps, MCP servers are called connectors. Anthropic keeps a directory of reviewed connectors at claude.ai/directory. You add them in your Claude settings.

Are MCP servers safe?

They are as safe as the server and the access you give it. Anthropic's docs say to verify you trust each server first, and warn that servers which fetch outside content can expose you to prompt injection. Start with read-only access.

What is the difference between MCP and a Claude skill?

MCP is a connection. It gives Claude a way into a tool. A skill is a set of instructions. It tells Claude how to do a job. You often use both: the server to reach the tool, the skill to use it your way.

Do I need to code to use an MCP server?

Not to use one. Adding a ready-made server is one command, or a few clicks in the Claude apps. Building your own server does take code, and that is a good project once you are comfortable.

See which MCP servers other builders use

Join the free Skool community. Ask what others have connected, and get help when a server will not load.