Fully accurate article, can confirm that it works really well. Don't need an MCP sever for Linear when the agent knows how to use the already there GraphQL API. The models already have decades of knowledge of how to use the Linux userspace including scripting languages like Python. Let them do what they're already good at.

MCP is a lack of trust that the model is smart enough to use the tools made for humans and that they need their own special thing. Nope, let em read the man page. They'll do fine.

This mirrors my experience.

If you want the agent to be more aware of the tool, prompt it to create a skill for whatever bash commands you wish