Claude Connectors: Which to Add and How to Add Your Own

Claude connectors are the integrations that let Claude read and act in your other apps, such as Gmail, Google Drive, Slack or Notion. You add them in Customize > Connectors, or paste in a server address to add a custom one. Every plan can add custom connectors (Free gets one), and Claude only sees what your own account can see.

Checked against Anthropic's help pages and connector directory on November 6, 2026. On that day the directory listed 1,083 connectors.

What are Claude connectors?

A connector is a link between Claude and one of your apps. It lets Claude look things up there and, if you allow it, take actions, like drafting an email or creating a task.

Under the hood each connector is an MCP server. MCP (Model Context Protocol) is a shared standard for plugging tools into AI assistants. An MCP server is the small service that tells Claude which actions an app offers and carries out the ones Claude picks.

The connector directory is the list of MCP servers Anthropic has reviewed. A custom connector is any other MCP server you add yourself by its web address.

Which Claude connectors are worth adding?

Add the two or three apps that hold the context for your most common task, and nothing else. On November 5, 2026, Anthropic's "Recommended" list opened with Google Drive, Gmail, Google Calendar, Canva, Microsoft 365, Notion, Figma, Slack, Atlassian, HubSpot, Asana and Linear, all marked "Anthropic verified".

One honest verdict: most people get more from one well-organised connector than from ten. Each connector you turn on adds to what Claude has to consider before it answers, and Anthropic itself suggests a different setting once you pass ten (more on that below).

How do I add a connector from the directory?

Open Customize > Connectors, pick the app and sign in. The full steps from Anthropic's help page:

1. Go to Customize > Connectors, or in a chat click + (or type /), then Connectors > Manage connectors.
2. Click + next to Connectors and browse or search the directory.
3. Open a connector to see what it can read and change.
4. Click Connect and sign in to that app.
5. Choose any settings it offers.

After that, turn it on or off for each chat from + > Connectors. On a Team or Enterprise plan, an owner has to switch a connector on for the organisation first. Each person still signs in with their own account.

Installing connectors from the phone app is still in beta, so add them on the web or desktop first. They then work on mobile too.

How do I add a custom connector to Claude?

Paste the server's web address into Customize > Connectors. Custom connectors work on Free, Pro, Max, Ultra, Team and Enterprise. Free accounts can add one. On Team, only owners can add them, and on Enterprise, an owner or a custom role with library access.

1. Go to Customize > Connectors and click Add, then choose the custom connector option. (Anthropic's two help pages label this button slightly differently, so look for "custom".)
2. Give it a name and paste the remote MCP server URL the tool gave you.
3. Claude checks the address and shows the sign-in method it found. Review it.
4. Pick how to sign in:
   • Sign in now
   • Sign in when needed
   • No sign in, for public servers
5. If asked, pick how Claude identifies itself to the server. "Use Claude's published identity" is the recommended default.
6. If the tool gave you an API key instead of a sign-in, add it under Request headers.
7. Click Add.

Two catches to know before you start:

• You can't edit a custom connector. To change the address or the sign-in, remove it and add it again.
• Claude reaches the server from Anthropic's cloud, not from your computer. That's true even in the desktop app and Cowork. A server that only runs on your laptop or behind a company VPN won't connect unless your IT team allows Anthropic's addresses through.

Full details are in Anthropic's custom connector guide.

What data can Claude access through connectors?

Only what your own account in that app can see, and only while the connector is on for the chat. Claude inherits your permissions: if you can't open a folder in Drive, neither can Claude. There's one exception. A custom connector that runs on a shared login reaches whatever that shared login can reach.

Three things people expect that aren't quite true:

1. "It reads everything." In practice it searches and reads a slice. Tiago Forte tested connectors on his own large accounts and found Claude read roughly 0.5–5% of a big source, leaning towards recent items (Forte Labs, June 2026). That's one person's tests, but it matches how search-based tools behave.
2. "It stays inside Anthropic." Connected apps process your data on their own systems, under their own terms.
3. "Anyone on my team can see what I connected." On Team and Enterprise, connectors only work in private projects, and chats that pull in synced content can't be shared.

Are Claude connectors safe?

They're as safe as the permissions you give them, and you can set those action by action. For each connector, go to Customize > Connectors, pick it, and open Tool permissions. Actions are grouped by kind, such as read-only and write or delete, and each can be set to Always allow, Needs approval or Blocked.

The risk to watch for has a name: the lethal trifecta. Here's how it works:

1. Claude can read your private data.
2. It reads something written by a stranger, like an incoming email or a web page.
3. It has a way to send things out.

A hidden instruction in step 2 can then ask Claude to send your data somewhere. Blocking or approving the "send" actions breaks the chain.

One setting removes that safety net. In Research mode, Claude can call connector tools without asking each time. Anthropic's advice is to switch off write actions before you start research.

Why does Claude get worse with lots of connectors?

Every active connector adds a list of actions Claude has to weigh, which takes up room in the conversation.

In a chat, open + > Connectors > Tool access. Auto is the default. Anthropic suggests switching to On demand once you have 10 or more connectors active, so Claude only pulls in a connector's actions when the task needs them.

Are Claude connectors the same as MCP servers?

Yes and no: every connector is an MCP server, but not every MCP server is one you'd add as a connector. A connector is an MCP server you've added to your Claude account, either from the reviewed directory or by its address.

MCP servers that run on your own computer plug into Claude Desktop or Claude Code instead. For how tools, MCP servers and skills fit together, see how tools, MCP and skills differ.

Do connectors work in Cowork, Claude Desktop and mobile?

Directory and custom connectors work on the web, in the desktop app, on mobile and in Cowork. Desktop extensions only work in the desktop app.

If you're setting up Cowork from scratch, our Cowork beginner's guide covers the rest of it. For the terminal version, see adding MCP servers to Claude Code.

Can one connector give Claude hundreds of apps?

Yes. A few services act as one connector with many apps behind it. You connect Claude once, then connect your apps inside that service.

The trade-off is one more company between Claude and your accounts, with its own pricing and its own terms. Zapier and Composio are about everyday apps. treg is mostly about data sources like search and company data. Tool routers compared goes deeper on choosing between them.

What if you'd rather not manage connectors at all?

Then you may not want connectors. You may want apps that come already connected. That's how Bento works.

Bento isn't a Claude connector, and it doesn't plug into Claude. It's a separate subscription where every plan already includes "Gmail, Slack, Notion and thousands more apps". You say what you want in a chat and it makes it: the weekly report sent, the post ready to share, the app at a link. Its agents browse the web and run on a schedule to get there.

The difference from the services above is who does the wiring. With Zapier or Composio you still connect each app to Claude through them and manage it there. In Bento the app connections come with the plan, all in one box: you sign in to each app once and the agents use it.

Plans start at $20 a month with a 7-day free trial. Message counts on Bento's plans are an estimate, not a cap.

When to stay put: if you live in Claude and need two or three connectors, add them in Claude. Bento makes sense when you'd rather hand over the job than wire up the tools for it.

Frequently Asked Questions

Are Claude connectors free?

Adding a connector costs nothing extra on any Claude plan, including Free. Free can use directory connectors and one custom connector. The connected app still bills you on its own plan. For example, Zapier counts two tasks for every action Claude runs through it.

How many connectors can I add?

Anthropic's help pages don't give a limit for directory connectors. Free accounts are limited to one custom connector. Past about ten active connectors, switch Tool access to On demand.

Can I edit a custom connector after adding it?

No. Remove it and add it again with the new address or sign-in.

Why won't my custom connector connect?

The most common reason is that the server isn't reachable from the public internet. Claude connects from Anthropic's cloud, not your computer. Check that the address works from outside your network, and that you picked the sign-in method the server expects.

Does ChatGPT have connectors too?

Yes, with a different setup. The ChatGPT version of this guide covers it.