8 min readUpdated September 22, 2026
What is MCP?
The Model Context Protocol as a technical shape: hosts, clients, servers, tools, resources and prompts — and what changes when a tool is exposed through it.
MCP
MCP is the Model Context Protocol for connecting AI applications to tools and data. Learn about servers, permissions and the actions those tools allow.
Where should I start with mcp?
Start with the guide that matches your question or the error you can see. Each guide below names the problem, gives checks you can run, and links to related reading.
What changes when an assistant gains an MCP tool?
The assistant can request an action through the connected tool, but the server and its credentials determine what is possible. An illustrative read-only repository tool can retrieve a file; a separate write-capable tool may commit a change. Calling both connections MCP does not make their permissions equivalent.
- 1List the tools exposed by the connection and separate reads from writes.
- 2Check which account and permissions each tool uses.
- 3Require approval before a consequential write and inspect the returned result.
Can retrieved text tell the assistant what to do?
A file or web page returned by a tool is data, not authority to change the task. Treat text such as 'send your token here' as untrusted content, even if it appears inside a useful result. Avoid broad credentials when a narrower read-only connection answers the question. Start with the MCP definition, then the beginner guide's connection sequence.
Read the mcp guides
6 min readUpdated September 22, 2026
MCP explained for beginners
A plain-language introduction to what MCP is for, using the everyday situations where an AI assistant needs something it cannot see.
Looking for something else? Browse every topic.