Main

Remote MCP is powerful but changes the trust model. Before adopting a hosted endpoint, decide:

• What data the MCP server can access (files, tokens, logs).
• Whether requests are logged and retained.
• How you pin versions or audit server updates.

For production, prefer self-hosting or an internal gateway with strict auth and allowlists.

README excerpt (verbatim)

Awesome Remote MCP Servers

📧 Subscribe to the Newsletter Stay informed about new MCP servers and updates: Subscribe

Not a developer & looking for ChatGPT Custom Integrations or Claude Custom Connectors? Check out the Remote MCP Server List

A curated, opinionated list of high-quality remote Model Context Protocol (MCP) servers. This repository focuses on listing official and well-maintained MCP servers that can be accessed over the internet, helping developers discover reliable and production-ready services for their AI applications.

What is MCP?

Model Context Protocol (MCP) is a protocol that enables AI models to interact with external tools and resources. Remote MCP servers allow clients to securely connect to these services over the internet.

What are Connectors?

This is OpenAI and Anthropic's branding for connecting to remote MCP servers.

What is the ChatGPT Apps SDK

These are just Remote MCP servers too! They additionally provide UI components as resources that can use displayed inside ChatGPT.

Why Remote only?

Remote MCP servers are a more secure and easier way to use MCP - compared to other methods of using MCP:

• Legitictacy: Servers can be evaluated by checking the domain in the URL - it should be a familiar and trusted domain.
• Permissioned: Servers should be assumed to have access to whatever data you've granted your agent access to, based on the authentication method used.
• Ease of use: You can copy & paste the server URL and you're away. No installing NPM packages.
• They are the ONLY way to use MCP servers with Web based clients.

How can I use an MCP server listed here?

There are two main ways to use remote MCP servers on this list.

1. Using a MCP ready client (Claude, ChatGPT, Cursor, etc.)
2. In an API request to an LLM provider (OpenAI, Anthropic, Gemini, etc.)

Authentication options

This list has a mix of authentication options, different methods are better suited to certain use cases:

• OAuth 2.1: This is an industry standard for authentication and is supported by MCP Clients like Claude. Adoption amongst clients is growing, but not yet widespread. For this all you need to connect is the server URL, and the client will guide you through the authentication flow.

FAQ

Q: Why remote MCP instead of local install? A: Speed and centralization: connect via URL and reuse infrastructure across machines.

Q: Is it safe by default? A: It depends on the endpoint’s operators; treat it like any remote service with auth/logging concerns.

Q: How do I install one? A: Pick a server URL from the list and add it to your MCP client using the HTTP transport instructions.