Guides
Reference guides on the Model Context Protocol and Claude Skills — what they are, how to use them, and how to judge what's safe to install. Written to stay accurate rather than to be news, so none of it carries a date.
Understanding MCP
Start here if the terminology is still fuzzy. These cover what MCP is and how it's put together.
- What is an MCP server?The plain-English definition: what a server exposes, how it reaches your AI client, and how it differs from an API or a plugin.
- MCP server architectureThe technical structure — host, client and server roles, capability negotiation, message flow, and the two standard transports.
- Local vs remote MCP serversWhere a server runs decides whether your data leaves your machine. The trade-offs in reach, trust and setup, side by side.
Using MCP servers
The practical side: getting a server running, and deciding whether you should.
- How to install an MCP serverWhere each client keeps its config file, what goes in the entry, and how to fix a server that won't connect.
- MCP server securityA server runs with your privileges. The real risks, what to check first, and what a security score can and can't tell you.
- How to build an MCP serverFor writing your own: the decisions to make first, what the protocol requires of you, and the mistakes that show up most.
Claude Skills
Skills are the other half of the picture — instructions rather than connections — and they're routinely confused with MCP.
- What are Claude Skills?What's inside a skill, how the three-stage loading model works, and why the description line matters more than the instructions.
- Claude Skills vs MCP serversSkills change how well something is done; MCP servers change what's reachable at all. Which one your problem actually needs.
- How to use Claude SkillsInstalling them, writing your own from a working template, and fixing a skill that never activates.