Understanding Connectors and MCP
Scorri per mostrare il menu
You've got your agent installed. Now you need to teach it how to reach into the apps you actually use.
A connector is a bridge between Claude and one of your apps. It's what lets Claude read your Gmail, edit your Google Docs, post to Slack, or add a row to a Notion database. Without connectors, Claude is just an AI in a window. With connectors, Claude can actually work inside your tools.
You might also hear the term MCP. It stands for Model Context Protocol. It's the open standard Anthropic built so AI tools can connect to apps in a consistent way. Think of MCP like USB: Gmail, Notion, Slack, and other tools can all plug into Claude through the same kind of protocol.
You don't need to understand MCP deeply. You just need to know that connectors and MCP are the same idea, just discussed at different levels.
Here's how to think about it in three layers:
When you connect Gmail to Claude, three things happen:
- Claude shows you what permissions it's asking for, such as "read my emails" or "send on my behalf."
- You log in to Gmail through a secure window.
- Gmail gives Claude a key that lets it act as you, within the permissions you approved.
Two important things to know:
- You can revoke a connector at any time. You can do this inside Claude settings or inside the connected app itself. Connectors are not permanent.
- A connector only gives Claude access to what you approved. If you did not grant "send email" permission, it cannot send emails. Permissions are granular.
Claude includes hundreds of prebuilt connectors. The big ones, such as Gmail, Google Calendar, Google Drive, Slack, Notion, Asana, Linear, Jira, and GitHub, work out of the box. In the next chapter, you'll connect Gmail, and the same flow works for every other connector.
Grazie per i tuoi commenti!
Chieda ad AI
Chieda ad AI
Chieda pure quello che desidera o provi una delle domande suggerite per iniziare la nostra conversazione
Understanding Connectors and MCP
You've got your agent installed. Now you need to teach it how to reach into the apps you actually use.
A connector is a bridge between Claude and one of your apps. It's what lets Claude read your Gmail, edit your Google Docs, post to Slack, or add a row to a Notion database. Without connectors, Claude is just an AI in a window. With connectors, Claude can actually work inside your tools.
You might also hear the term MCP. It stands for Model Context Protocol. It's the open standard Anthropic built so AI tools can connect to apps in a consistent way. Think of MCP like USB: Gmail, Notion, Slack, and other tools can all plug into Claude through the same kind of protocol.
You don't need to understand MCP deeply. You just need to know that connectors and MCP are the same idea, just discussed at different levels.
Here's how to think about it in three layers:
When you connect Gmail to Claude, three things happen:
- Claude shows you what permissions it's asking for, such as "read my emails" or "send on my behalf."
- You log in to Gmail through a secure window.
- Gmail gives Claude a key that lets it act as you, within the permissions you approved.
Two important things to know:
- You can revoke a connector at any time. You can do this inside Claude settings or inside the connected app itself. Connectors are not permanent.
- A connector only gives Claude access to what you approved. If you did not grant "send email" permission, it cannot send emails. Permissions are granular.
Claude includes hundreds of prebuilt connectors. The big ones, such as Gmail, Google Calendar, Google Drive, Slack, Notion, Asana, Linear, Jira, and GitHub, work out of the box. In the next chapter, you'll connect Gmail, and the same flow works for every other connector.
Grazie per i tuoi commenti!
