Get started with custom connectors using remote MCP (original) (raw)

What are custom connectors?

Custom connectors let you connect Claude directly to the tools and data sources that matter most to your workflows. This enables Claude to operate within your favorite software and draw insights from the complete context of your external tools.

You can:

What are remote MCP servers?

The Model Context Protocol (MCP) is an open standard, created by Anthropic, for AI applications to connect to tools and data.

Previously, MCP servers only ran locally (i.e. on a user's laptop). Now, developers can build and host remote MCP servers that communicate with AI apps over the internet.

Remote MCP servers give models access to internet-hosted tools and data, transforming Claude into an informed teammate that can independently handle complex, multi-step projects tailored to your needs.

Network requirements

When you add a custom connector, Claude connects to your remote MCP server from Anthropic's cloud infrastructure, rather than from your local device. This is true across every Claude client, including claude.ai, Claude Desktop, Cowork, and the mobile apps.

This means your MCP server must be reachable over the public internet from Anthropic's IP ranges. Servers hosted on a private corporate network, behind a VPN, or blocked by a firewall won't connect, even if you can reach them from your own machine.

If your server is on a private network

You’ll need to allowlist Anthropic's IP addresses in your firewall so inbound connections from Claude can reach your server. See Anthropic IP addresses for the current ranges.

Why this applies to Cowork and Claude Desktop

Even though Cowork and Claude Desktop run on your computer, remote connectors are configured and brokered through your Claude account. The connection to your MCP server originates from Anthropic's servers, not from your machine's network interface. Local MCP servers configured in Claude Desktop via claude_desktop_config.json are a separate mechanism and do use your local network, but those aren't available in Cowork or claude.ai.

Add a custom connector

For Team and Enterprise plans

Preliminary steps for owners:

Before members of Team and Enterprise plans can configure custom connectors, an Owner or Primary Owner needs to follow these initial steps to add a custom connector to your organization:

Steps for members after connector is configured:

For Pro and Max plans

If you are using an individual Pro or Max plan, follow these steps to add a custom connector:

Enabling connectors after configuration

You can enable connectors for individual conversations via the “+” button on the lower left of your chat interface, then "Connectors." You'll see your configured connectors with toggles allowing you to enable/disable them per conversation.

Remove custom connectors

You can remove a custom connector by following these steps:

If you're hoping to edit a custom connector, you'll need to remove it first, then re-add it using the updated details.

Build custom connectors

Security and privacy considerations

Custom connectors allow you to connect Claude to arbitrary services that have not been verified by Anthropic. When you connect Claude to external services, you're granting it the ability to access and potentially modify data within those services based on your permissions. It’s important to make sure you’re only connecting to remote MCP servers that you trust and that you’re aware of Claude’s interactions with web connectors.

Security and permissions

When you add a custom connector to Claude, you'll typically go through an OAuth authentication process to securely sign in to the application and grant specific permissions. This allows Claude to interact with the application on your behalf, without Claude ever seeing your actual password. You can revoke these permissions at any time by disconnecting the connector in Claude's settings or the third-party service's security settings.

Remote MCP servers act as intermediaries between Claude and external applications. You should:

Reporting malicious MCP servers

Taking actions with tools

Remote MCP servers give Claude tools it can invoke during your conversation. The developer of an MCP server can define what these tools do, including:

Claude can only access resources that you've given the server permission to access, but you should:

Interactive connectors

Some connectors can display interactive interfaces directly within your Claude conversations. Instead of only returning text-based responses, these connectors can open live, interactive apps — like dashboards, task boards, or design tools — right in the chat.

Interactive connectors appear in two ways:

You can interact with these connectors directly — filtering data, checking off tasks, adjusting settings — without leaving the conversation. Any actions you take within the interface use the same permissions you granted when connecting the tool.

Admin controls: Team and Enterprise plan owners can disable specific tool calls that render interactive connectors within Organization settings > Connectors.

Using Claude with Research

Research allows Claude to deeply investigate queries by searching through hundreds of internal and external sources. During the research process, Claude can invoke tools from your connectors automatically without further approval.

When using Research with custom connectors:

Use connectors to extend Claude's capabilities When to use desktop and web connectors Use interactive connectors in Claude MCP connectors MCP: Individual connectors