Topics

What Is an MCP Server?

An MCP server is a small service that gives an AI client structured access to tools, data, or actions through the Model Context Protocol.

Browse MCP serversSee examples

Short version

An MCP server is a bridge between an AI client and an outside capability. The client sends structured requests, the server performs a tool call or returns context, and the result comes back in a format the agent can use.

Step 1
Client asks
Step 2
Server executes
Step 3
Agent uses result

How MCP servers work

Tools
Actions the agent can call, such as searching docs, opening a browser, querying a database, or creating a ticket.
Resources
Structured context the client can read, such as files, docs, schemas, records, or remote data.
Prompts
Reusable instructions or workflows that help the client use a server in a consistent way.

Common MCP server examples

Browser automation
A browser automation server can let an agent open pages, inspect UI state, and run browser actions.
Documentation lookup
A docs-focused server can retrieve current package context before an agent writes code.
Database context
A SQL-focused server can expose schemas, query helpers, and project metadata to a coding agent.

When to use one

  • Your agent needs reliable access to a tool or data source.
  • You want reusable setup instead of pasting context into every prompt.
  • The workflow benefits from structured results instead of unstructured text.

What to watch

  • Scope credentials so a server can only access what it needs.
  • Test new servers with low-risk workflows before using them on sensitive data.
  • Avoid adding many overlapping servers to one client without a clear reason.

MCP server FAQ

What does MCP stand for?

MCP stands for Model Context Protocol. It defines a standard way for AI clients to connect to tools and context providers.

What is the difference between an MCP server and an MCP client?

An MCP client is the AI application you use, such as Claude Desktop or a coding agent. An MCP server is the service that exposes tools, data, or actions to that client.

Do MCP servers run locally or remotely?

Both are possible. Some MCP servers run as local commands on your machine, while others expose remote HTTP or SSE endpoints.

Are MCP servers safe to use?

They can be safe when permissions are scoped carefully. Treat them like any integration that can read files, access data, browse websites, or trigger actions.