What is MCP, and what is it useful for?
Not in your language? Paste the link or the text into your AI and let it read it for you.
How to do it
Click an avatar to see the source post- 1
MCP can be understood as a unified socket standard for AI to connect to external tools and data, specifying how AI applications plug into external services such as GitHub and databases.

- 2
Once MCP is connected, agents can directly call tool functions within a conversation. For example, the REA MCP service can be registered to multiple agents with a single command: npx rea-agents setup.

- 3
Adding Serena MCP to Codex lets it pinpoint code by function, class, and reference without re-reading entire files, and the author says this noticeably reduces token consumption.

- 4
There is also an MCP for Blender. Someone hooked it up to Codex, used concept art as a reference, and reportedly made a 3D game in 45 minutes.

- 5
More MCPs isn't always better. The more tools like browsers, Figma, and GitHub you have, the greater the context burden, so load only what the current project needs.

- 6
The MCP protocol itself is still being updated. The latest version switches to stateless request-response, where each request is independent, and servers can be deployed like ordinary HTTP services.

What people on X saySorted by usefulness
Everything here comes from real-world tests by users on X. Rules and products change often, so always check the latest. If you know a better way, we'd love to hear it.