How to connect
an MCP server
MCP stands for Model Context Protocol. It lets AI applications use tools and data through MCP servers. To connect one, you need an application that supports MCP and the server's setup instructions.
Your AI application
The MCP client: the place you ask questions and use tools.
An MCP server
The bridge that exposes a set of tools, resources or prompts.
Your tools & data
The files, services and systems you want to work with.
Your first connection
- 1
Start with one job
Choose something concrete: search your notes, inspect a database, or test a page in a browser. A small, specific task makes choosing the right server much easier.
Find a server - 2
Read the project’s setup docs
Check its supported client, operating system and runtime. Some servers run locally; others expose a remote endpoint. Look for the actual installation and authentication instructions rather than guessing from a badge.
Explore developer tools - 3
Connect through your MCP client
Your AI application is the client. Follow its documentation to add the server’s local command or remote URL. Use the credentials the project requires, with access limited to what your task needs.
Understand the connection - 4
Try a small request
Ask your assistant to do the task you picked. Review any tool-use permission prompts and check the result against the original resource. Add more tools when you have a reason to.
Explore browser tools
A few things the badges don’t tell you
“Cloud” describes a connected service, not a guaranteed remote MCP endpoint. “Official” comes from the community list, not a security audit. Popularity is useful context; read the permissions and setup docs before connecting.
Read the official MCP introduction