---
title: "MCP Setup Guides: Claude, Cursor, VS Code, Codex, ChatGPT"
url: "https://mcpshelf.co/guides"
description: "How to add a server in each app, how to test it, and how to fix it when it will not start."
site: "MCP Shelf"
---

# Setup guides

How to add a server in each app, how to test it, and how to fix it when it will not start.

## Set up your app

Where servers go in each app, with configs you can copy.

- [How to add MCP to Claude Code](https://mcpshelf.co/guides/claude-code): To add MCP to Claude Code, run one command: claude mcp add. You never have to open a JSON file unless you want to share servers with your team. Here is the short version, then the details that catch people out.
- [Adding MCP servers to Claude Desktop](https://mcpshelf.co/guides/claude-desktop): Claude Desktop reads local MCP servers from claude_desktop_config.json. Remote servers can also be added as connectors in Settings without touching the file.
- [Adding MCP servers to Cursor](https://mcpshelf.co/guides/cursor): Cursor reads MCP servers from a file called mcp.json, and lists them on its Customize page, where you can switch each one on or off. You can add servers from there or edit the file directly; both end up in the same place.
- [Where Cursor keeps mcp.json](https://mcpshelf.co/guides/cursor-mcp-json-location): There are two mcp.json files. One lives in your home folder and applies everywhere; the other lives inside a project.
- [Adding MCP servers to VS Code](https://mcpshelf.co/guides/vs-code): VS Code gives MCP tools to GitHub Copilot in agent mode. You add a server from the Command Palette or by editing mcp.json, and the one thing that trips everyone up is the key name: VS Code's own file uses "servers", not "mcpServers".
- [Adding MCP servers to Codex](https://mcpshelf.co/guides/codex): Codex keeps MCP servers in config.toml, written in TOML rather than the JSON most READMEs show. The Codex CLI, the IDE extension and the ChatGPT desktop app share that file, so you set a server up once. Here is how to add one, and how to translate any JSON config.
- [Using MCP servers in ChatGPT](https://mcpshelf.co/guides/chatgpt): ChatGPT on the web does not run programs on your computer, so it can only use MCP servers that someone hosts at a URL. You add one as a custom plugin. For a server that only gives you an npx or uvx command, use the ChatGPT desktop app or Codex, which share a config, or another desktop app.
- [Adding custom connectors to Claude](https://mcpshelf.co/guides/claude-connectors): A custom connector is a remote MCP server you add to your Claude account by its URL. Add it once on claude.ai and it follows you to Claude Desktop and the mobile apps, with no config file to edit.
- [Adding MCP servers to JetBrains IDEs](https://mcpshelf.co/guides/jetbrains): JetBrains IDEs such as IntelliJ IDEA, PyCharm and WebStorm connect to MCP servers through AI Assistant. You paste the same "mcpServers" JSON other apps use into a settings dialog, and each server can be on for every project or just one.

## How MCP works

What a server is, and the two ways to run one.

- [What is an MCP server?](https://mcpshelf.co/guides/what-is-an-mcp-server): An MCP server is a small program that gives an AI assistant something it cannot do on its own: read your files, query a database, drive a browser, file a Jira ticket. MCP (Model Context Protocol) is the shared plug that lets any assistant use any of these servers.
- [Remote or local: two ways to run an MCP server](https://mcpshelf.co/guides/remote-vs-local-mcp-servers): Every server on MCP Shelf runs one of two ways, and some run both. The choice decides who holds your data and what you have to install.
- [mcp-remote: use a remote MCP server in any app](https://mcpshelf.co/guides/mcp-remote): mcp-remote is a small npm package that your app starts like any local server. It connects to a remote MCP server for you, handles the browser sign-in, and passes the messages back and forth. You need it when an app or config file only accepts a command.

## Build and run your own

Write a server of your own, and run servers in Docker.

- [Running MCP servers in Docker](https://mcpshelf.co/guides/docker-mcp-servers): Docker runs an MCP server in its own container, so it cannot see the rest of your computer unless you let it, and you need nothing installed but Docker. You can put a `docker run` command straight into your app, or let Docker Desktop’s MCP Toolkit manage a set of servers behind one gateway.
- [How to build your own MCP server](https://mcpshelf.co/guides/build-an-mcp-server): An MCP server is a small program that tells an AI app which tools it offers, then runs them when asked. With the official SDKs, a working server with two tools is about 40 lines of TypeScript or Python. This guide builds one, tests it, and plugs it into your app.

## Test, fix and stay safe

Prove a server works, fix the errors, and check it before you trust it.

- [Testing an MCP server with MCP Inspector](https://mcpshelf.co/guides/test-an-mcp-server): When a server does not work in your app, you need to know whether the server or the app is at fault. MCP Inspector, the official testing tool, connects to one server on its own and shows you exactly what it offers and what goes wrong. It takes about a minute.
- [Fixing MCP server errors](https://mcpshelf.co/guides/mcp-server-errors): Most MCP problems come down to five things: the app cannot find the command, the server crashes on start, it starts too slowly, the config has a typo, or the sign-in failed. Find the real error message first, then look it up below.
- [Checking an MCP server before you install it](https://mcpshelf.co/guides/check-an-mcp-server): A local MCP server is a program running as you, with your files, your network and whatever tokens you give it. An assistant decides when to call it. That is worth a few minutes of checking.

## Where each app keeps its servers

| App | Where servers go | Format | Remote URLs | Local commands |
| --- | --- | --- | --- | --- |
| [Claude Code](https://mcpshelf.co/guides/claude-code) | `claude mcp add`, or `.mcp.json` in a project | "mcpServers" | Yes, with "type": "http" | Yes |
| [Claude Desktop](https://mcpshelf.co/guides/claude-desktop) | `claude_desktop_config.json`, or Customize → Connectors | "mcpServers" | As connectors | Yes |
| [Cursor](https://mcpshelf.co/guides/cursor) | `~/.cursor/mcp.json` or `.cursor/mcp.json` | "mcpServers" | Yes | Yes |
| [VS Code](https://mcpshelf.co/guides/vs-code) | `.vscode/mcp.json`, or MCP: Add Server | "servers", each with a "type" | Yes | Yes |
| [Codex](https://mcpshelf.co/guides/codex) | `~/.codex/config.toml`, or `codex mcp add` | TOML: [mcp_servers.NAME] | Yes (Streamable HTTP) | Yes |
| Windsurf | `~/.codeium/windsurf/mcp_config.json` | "mcpServers", "serverUrl" for URLs | Yes | Yes |
| [Claude on the web](https://mcpshelf.co/guides/claude-connectors) | Customize → Connectors → add a custom connector | A URL, no file | Yes, as connectors | No |
| [JetBrains IDEs](https://mcpshelf.co/guides/jetbrains) | Settings → Tools → AI Assistant → Model Context Protocol (MCP) | "mcpServers" | Yes | Yes |
| [ChatGPT on the web](https://mcpshelf.co/guides/chatgpt) | chatgpt.com/plugins | A URL, no file | Yes, as plugins | No |

---

From [MCP Shelf](https://mcpshelf.co/): A short shelf of MCP servers that are worth installing. Index for agents: https://mcpshelf.co/llms.txt

