How to connect Claude to your apps with connectors and MCP
Claude connectors are integrations, built on the Model Context Protocol (MCP), that let Claude read from and act in other apps such as Google Drive, Gmail, Slack, GitHub, Notion and CRMs. You can add reviewed connectors from Claude’s Connectors Directory or add a custom connector by pasting a remote MCP server URL. Custom connectors work on all plans, with free users limited to one; on Team and Enterprise, owners add connectors for the organisation.
What are Claude connectors?
A connector lets Claude use another app during a conversation. Under the hood, each connector is an MCP server: a small service that describes the tools and data it offers in a standard way. Once connected, Claude can search your files, read messages, look up records or take actions such as creating a ticket, using the permissions you grant. Our beginner’s guide to MCP explains the protocol itself.
What is in the Connectors Directory?
The directory lists connectors reviewed by Anthropic, grouped by category such as productivity, development, sales and design. It grew quickly in 2026: one count found 369 entries across 11 categories in June, up from around 50 at the start of the year, according to DirJournal’s connector list. Counts change weekly, so check the directory for what is available now.
How do you add a connector?
- Open connector settings: In Claude, go to Customise, then Connectors.
- Browse or add custom: Pick a connector from the directory, or choose “Add custom connector”.
- Enter the server URL: For a custom connector, paste the remote MCP server’s HTTPS address.
- Authenticate: Sign in to the app and approve the permissions requested.
- Test it: Ask Claude a question that needs the connected app, such as “find last week’s proposal in Drive”.
Setup steps follow Anthropic’s help centre. Custom connectors using remote MCP are available on Free, Pro, Max, Team and Enterprise plans, and free users are limited to one.
How do connectors work on team plans?
| Plan | How connectors are added |
|---|
| Free | Directory access varies; one custom connector |
| Pro and Max | Directory connectors and multiple custom connectors |
| Team and Enterprise | Owners add connectors for the organisation; each user then signs in with their own account |
| Enterprise extras | Admin policies, audit logging and controls over which connectors are allowed |
What are useful first connectors?
- File storage: Google Drive, OneDrive or SharePoint for briefs and reports.
- Communication: Slack or Gmail for summaries and drafts.
- Development: GitHub or Linear for issues and code.
- Sales: HubSpot or Salesforce for account research.
- Design: Figma for design files and specs.
- Automation: Zapier or n8n to reach apps without their own connector.
Automation platforms act as a bridge to thousands of apps; see our comparison of n8n, Zapier and Make.
How do you keep connectors safe?
- Prefer directory connectors or servers your team built and reviewed
- Start read-only and add write actions deliberately
- Approve actions that send messages, change records or delete anything
- Watch for prompt injection: content in connected apps can contain hidden instructions
- Remove connectors nobody uses
Our guide to human-in-the-loop AI explains where approvals belong.
What does this look like in practice?
A sales lead connects HubSpot and Gmail, then asks Claude: “Prepare a one-page brief for tomorrow’s call with Acme, including open deals and our last three emails.” Claude searches the CRM, reads the thread and writes the brief in one step. A product manager connects Linear and Slack and asks for a summary of everything said about a feature this week, with open bugs listed. A designer connects Figma and asks Claude to turn a frame into written specs for developers. In each case, the person stays in the conversation and reviews the result before acting on it.
What if a connector does not work?
- Check you signed in with the right account and approved the permissions
- Confirm your plan supports the connector, and on Team plans that an owner has enabled it
- For custom connectors, make sure the server URL is public and uses HTTPS
- Disconnect and reconnect to refresh expired authentication
- Ask Claude which tools it can see, to confirm the connector loaded
Combining connectors with Claude Skills and Claude Projects gives Claude both the method and the context for recurring work.
What if there is no connector for your tool?
You can build a remote MCP server for an internal system or a niche app, then add it as a custom connector. Teams on Team or Enterprise plans can also submit connectors to the directory. If you only need browser actions, Claude in Chrome can work directly in web apps without a connector.
To understand what sits behind each connector, read MCP servers explained; for the risks, see MCP security risks.
How UIUX Studio helps
We build custom MCP servers and connectors for internal systems and help teams set up safe connector policies. Talk to our Agentic AI & Automation and product engineering teams.
Frequently asked questions
Are Claude connectors the same as MCP servers?
Usually yes; a connector is typically a remote MCP server Claude can reach over HTTPS.
Can free users add connectors?
Free users can add one custom connector.
Who adds connectors on Team plans?
Owners add them for the organisation; each user then authenticates with their own account.
Can Claude change data through connectors?
Yes, if the connector offers write tools and you grant permission.
Is it safe?
It can be, with vetted connectors, minimal permissions and approvals for actions.
Related articles