Glossary
What is an MCP server?
An MCP server is a program that exposes tools, data or prompts to AI applications using the Model Context Protocol. It might wrap an API (create an invoice), a database (run a query) or a local resource (read files). AI clients connect to it and the model calls its tools when needed.
What it exposes
- Tools: functions with defined inputs, like search_orders(customer).
- Resources: readable data like files or records.
- Prompts: templates the user can invoke.
Local vs remote
| Local | Remote | |
|---|---|---|
| Transport | Standard input/output | HTTP |
| Runs on | The user's machine | A server |
| Auth | Local credentials | OAuth or API keys |
| Install | Command in client config | URL |
Publishing one
Publish the code or package, write clear tool descriptions (models choose tools based on them), document setup, then list it on MCP registries. See our guides for Smithery and Glama.
Example
A SaaS publishes a remote MCP server so customers can ask their AI assistant to pull reports from it. Listing it in three registries brings developer sign-ups who found it while browsing analytics servers.
FAQ
Do I need to code to use an MCP server?
No; many install with a command or URL in the AI app's settings.
Can an MCP server be malicious?
Yes, like any software; install trusted servers and review permissions.
How do models know when to use a tool?
From the tool names and descriptions the server provides.
How do I install an MCP server?
Add it to your AI app's MCP settings with a command (local) or URL (remote), following the server's docs.