WhyVault
Sign indeStart for free

Claude connectors

Claude connectors: your tools in every chat.

With connectors, Claude reads your notes, files or tickets instead of you pasting everything in. Here’s what connectors are, how to add a custom one in five minutes, and which ones are actually worth it – honestly, and without a top-ten list.

Reading time: 8 minutesUpdated:

01

What connectors in Claude are

A connector is a link between Claude and another service. When it’s switched on, Claude can use tools there: search, read, sometimes write. Instead of uploading a file or pasting in text, you just say “check my notes”, and Claude fetches what it needs.

Under the hood it’s MCP, the Model Context Protocol. The service runs an MCP server; Claude is the client. Because MCP is an open standard, a server built once also works with ChatGPT, Cursor or Claude Code. What is MCP?

There are two ways to get a connector. Claude’s directory lists connectors Anthropic has reviewed; you click connect and sign in. A custom connector is any other remote MCP server: you enter its URL, and Claude connects to it directly.

According to Anthropic’s help pages, custom connectors work on every plan, including Free, where you get one. On Team and Enterprise, an Owner adds the connector for the organisation. What exactly is available depends on your plan and can change; when in doubt, Anthropic’s current help page wins.

02

Directory connector or custom connector?

Both use MCP. The difference is who reviews it and how you connect.

–From the directoryCustom connector
Adding itPick it in the directory and connectEnter a name and the MCP server’s URL
Reviewed by AnthropicYesNo, you decide whom to trust
Which servicesWhatever is in the directoryAny remote MCP server, including small or self-built services
Sign-inUsually OAuthUsually OAuth; some servers need their own credentials
Free planDepends on your plan and the connectorOne custom connector
Team and EnterpriseOwner enables it, members connectOwner adds it, members click Connect
Adding it
From the directoryPick it in the directory and connect
Custom connectorEnter a name and the MCP server’s URL
Reviewed by Anthropic
From the directoryYes
Custom connectorNo, you decide whom to trust
Which services
From the directoryWhatever is in the directory
Custom connectorAny remote MCP server, including small or self-built services
Sign-in
From the directoryUsually OAuth
Custom connectorUsually OAuth; some servers need their own credentials
Free plan
From the directoryDepends on your plan and the connector
Custom connectorOne custom connector
Team and Enterprise
From the directoryOwner enables it, members connect
Custom connectorOwner adds it, members click Connect

If a service isn’t in the directory but offers an MCP URL, a custom connector is the direct route. That’s how you add WhyVault, for example.

03

Step by step: add a custom connector

Using WhyVault, your memory for every agent, as the example. Other servers work the same way, just with their URL.

  1. 1

    Open connectors

    Open Claude in the browser or the desktop app and go to Customize → Connectors. Click +, then Add custom connector.

  2. 2

    Enter a name and the URL

    Name it WhyVault and use the MCP URL as the address. If Claude asks about the OAuth client, choose Register automatically, not “Use Claude’s published identity”. Click Add.

    https://mcp.whyvault.io
  3. 3

    Sign in and approve

    Claude opens the WhyVault sign-in window. Sign in and approve access. Claude can then read and write, and delete only if you allow it. No access key needed.

  4. 4

    Switch it on in a chat

    Connectors are switched on per chat. In the chat, click + at the bottom left → Connectors and turn WhyVault on. Try: “Read my vault and summarise what I’m working on right now.”

  5. 5

    Check the permissions

    In WhyVault, Connect agents shows what the connection may do: read, write, delete. Deleting stays off until you turn it on. In Claude, you set per tool whether it asks first. Allow deleting

04

What the first chat looks like

With the connector on, Claude fetches the context itself:

Claude

  1. You: Read my vault and summarise what I’m working on right now.

  2. Tool: vault_overview · search_notes

    Claude: You have three active rooms: Work, Ventures and Private. In Ventures, the newsletter launch has been on since last week, with an open decision about the publishing day. In Work, the client project runs until the end of the month.

  3. You: Note it down: the newsletter goes out on Thursdays.

  4. Tool: append_to_note · “Newsletter – launch”

    Claude: Done. I added the decision with its reason under “Decisions”. You’ll see the change in your history and can undo it there.

Claude starts with `vault_overview`, then searches for what it needs instead of loading everything.

05

Which connectors are worth it

Not a ranking, just a rule of thumb: a connector pays off when you’d otherwise keep explaining or pasting the same things into Claude.

  • Memory and notes

    The biggest win, because every chat benefits: decisions, projects and preferences Claude doesn’t need to hear again. Best kept somewhere your other assistants can read too. Claude memory compared

  • Files and drives

    Useful if your documents already live in cloud storage. Claude finds the right file instead of you hunting for it and uploading it. Check what the connector gets access to: often it’s the whole drive.

  • Calendar and email

    Handy for questions like “When am I free next week?” or “What did the client ask for again?”. Also the most sensitive category: inboxes are full of other people’s text and personal details. Start with read-only.

  • Code and GitHub

    Worth it if you talk about code in Claude: issues, pull requests, files in a repo. If you code a lot, Claude Code is often the better fit, since it adds MCP servers directly.

  • Project management

    Makes sense if your tasks live in one tool and you have Claude write status reports or plans. Hold off on write access until you’ve seen how Claude handles your tasks.

06

Security: what to watch for

A connector gives Claude access to a real service. That’s the point, and it’s why a quick check is worth it:

  • Only trusted servers. Anthropic doesn’t review custom connectors. Only connect servers from providers you’d trust with your data anyway.
  • Keep permissions tight. Read what the server asks for when you sign in. Whatever Claude doesn’t need, deny it or switch it off later.
  • Watch for hidden instructions. Content from a connector, such as an email or a web page, can contain text meant to steer Claude (prompt injection). That’s why WhyVault marks note content as data, not instructions. Still, stay alert if Claude suddenly wants to do things you didn’t ask for.
  • Approve write tools deliberately. Have Claude ask before tools that change or send something, and use “Allow always” only for servers you trust.
  • Only switch on what the chat needs. Connectors are toggled per chat. What’s off can’t read anything.
  • Keep track of changes. In WhyVault, every change shows up in the history with the agent’s name, undo in one click. You revoke access under Connect agents. History and control

07

When it doesn’t work: the usual suspects

The full list is in the Claude setup guide. The short version:

  • “Add custom connector” is greyed out: on the free plan you already have a custom connector, or you’re a member on Team or Enterprise. Then the Owner adds it.
  • Sign-in fails right after “Add”: probably “Use Claude’s published identity” is selected. Remove the connector and add it again with Register automatically.
  • The sign-in window doesn’t open: allow pop-ups for claude.ai. In the desktop app, the sign-in opens in your default browser.
  • The connector doesn’t appear in the chat: it’s added but not switched on. + → Connectors → turn it on.
  • Claude can read but not write: the connection is set to read only in WhyVault. Change the permissions, then disconnect and reconnect so Claude reloads the tool list.
  • The connector shows “Disconnected”: access was revoked or the sign-in expired. Click Connect and approve again.

The best connector is the one that means you never explain anything to Claude twice.

?

Frequently asked questions

What are Claude connectors?

Links between Claude and other services, such as notes, files or a calendar. Technically they run over MCP: the service provides an MCP server, and Claude uses its tools when you switch the connector on in a chat.

What is a custom connector?

A connector you add yourself with the URL of a remote MCP server, instead of picking it from Claude’s directory. That’s how you connect services that aren’t listed there, for example WhyVault with https://mcp.whyvault.io.

Do custom connectors work on the free Claude plan?

According to Anthropic, yes, with one custom connector. Paid plans allow more than one. Plans change from time to time; when in doubt, Anthropic’s current help page wins.

Do connectors work in the Claude mobile app?

Yes. You add the connector on the web or in the desktop app, and it’s then available in the app on iOS and Android. You don’t have to set it up again there.

Are connectors safe?

As safe as the server behind them and the permissions you grant. Only connect trusted servers, grant as little as needed and have write tools ask first. Content from connectors can contain hidden instructions; Claude shouldn’t follow them, but stay alert.

Do ChatGPT and Claude Code use the same connection?

No, each assistant connects on its own. The URL is the same, though: with https://mcp.whyvault.io, Claude, ChatGPT, Cursor and Claude Code all work in the same vault. All agents

Where does WhyVault store my data?

WhyVault is EU-hosted (Frankfurt) and GDPR-compliant. What Claude reads from your vault is also processed by Anthropic. Claude, ChatGPT and your data

All guides

Your first connector: your memory.

Add WhyVault as a custom connector in five minutes, then use the same URL in ChatGPT and Cursor. Free up to 20 notes, no credit card.

Start for free