# Remote vs local MCP servers: what is the difference?

A local MCP server is a program the AI app starts on your own computer and talks to over standard input and output (stdio). A remote MCP server runs on someone else's machine and the app reaches it over the internet by URL, using Streamable HTTP and usually a sign-in.

One server page says "runs locally", another says "hosted". The difference decides where the server runs, what it can reach, and which assistants can use it.

*Diagram: Local and remote MCP servers.* A local server runs on your computer and talks to the app over stdio. A remote server runs on the internet; the app reaches it by https URL, usually after you sign in.

## Local servers

- The app starts the server itself, from a command in its config such as `npx -y @playwright/mcp`.
- They talk over stdio, so nothing listens on the network.
- The server runs with your user account's permissions: it can read and change what you can. Only run local servers you trust.
- You need the server's runtime installed (often Node.js for `npx`, or Python with `uvx`).

## Remote servers

- Someone hosts the server; you add it by URL, such as `https://mcp.example.com/mcp`.
- They use Streamable HTTP (the older SSE transport still appears; Claude Code's docs call it deprecated).
- Most ask you to sign in with OAuth or to add an API key in a header.
- Nothing to install, and they work in apps that can't start programs on your computer.

## Which apps take which

What each app's own documentation says, as of 19 November 2026:

- **Claude Desktop**: local servers in `claude_desktop_config.json`; remote servers as custom connectors.
- **Claude Code**: both, with `claude mcp add`.
- **Cursor**: both, in `mcp.json`.
- **VS Code**: both, in `mcp.json` or with the MCP: Add Server command.
- **ChatGPT**: remote servers over SSE or streaming HTTP. Running a server on your computer isn't documented; the setup offers a Server URL or a Secure MCP Tunnel.
- **Grok Bot**: remote servers as Remote HTTPS; its Command type runs a server on the Bot's cloud computer, not on yours.

Remote connectors in Claude and ChatGPT are called from the provider's cloud, not from your computer, so a server only reachable on your own network won't work there unless the provider offers a tunnel or you allow its addresses.

## In the directory

Each [MCP server](https://ai-powered.com/mcp) page says whether the server runs locally, is hosted, or both, and hosted servers show the results of our latest checks.

## Sources

Official documentation, checked on 19 November 2026:

- [MCP architecture overview (modelcontextprotocol.io)](https://modelcontextprotocol.io/docs/learn/architecture)
- [Custom connectors using remote MCP (Claude Help Center)](https://support.claude.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp)
- [Add custom MCP server (OpenAI developer docs)](https://developers.openai.com/api/docs/guides/developer-mode)
- [Computer and apps (xAI Grok Bot docs)](https://docs.x.ai/grok-bot/computer-and-apps)

## Related terms

- [MCP](https://ai-powered.com/academy/glossary/mcp): MCP (Model Context Protocol) is an open standard for connecting AI applications to outside tools and data. An AI app connects to small programs called MCP servers, asks each one what it offers, and calls those tools when a task needs them.
- [MCP server vs API](https://ai-powered.com/academy/glossary/mcp-server-vs-api): An API is an interface that a program calls with code written for that API. An MCP server offers its capabilities through one shared protocol, describing each tool so that any MCP-compatible AI app can discover and call it without custom code. Many MCP servers are a thin layer over an existing API.
- [Tool call](https://ai-powered.com/academy/glossary/tool-call): A tool call is a structured request from an AI model to run one of the functions it was given, with arguments that match the function's input schema. The app runs the function, sends the result back to the model, and the model uses it to answer or to decide its next step.
