Claude MCP: Which Business Tools Can Claude Connect To?
Which business tools Claude can connect to today, when you need a custom connector, and what to check before you connect anything.
Claude MCP connectors let Claude read from, and sometimes act in, the tools you already use. That might be your email, your calendar, your shared drive or your CRM. The question for an owner is simple: which of my tools can Claude actually reach today?
This article is the practical follow-on to What is MCP?, so I will not explain the idea again. Instead, it covers real connectors, what to check before you switch one on, and what you can ask once it is working. The official directory and help pages were checked on 12 November 2026.
Which business tools Claude can connect to today
Anthropic keeps a connectors directory. It describes connectors as links between Claude and the tools you already use, built for the Model Context Protocol.
On the day I checked, the directory listed these tools, among others:
- Google Drive, Gmail and Google Calendar
- Microsoft 365
- Slack and Notion
- HubSpot, for customer records
- Asana, Linear and monday.com, for tasks and projects
- Atlassian, Intercom and Shopify
- Box, Canva, Figma and Adobe
Some of these suit a developer more than an owner, such as Supabase and Vercel. The directory changes, so treat my list as a snapshot and look at the live page.
Anthropic's Team plan help page says Team includes connectors for Google Drive, Gmail, Google Calendar, GitHub, Microsoft 365 and Slack.
How Claude MCP connectors work day to day
Per the connectors help page, you open the directory, pick a connector, click Connect and sign in. Then you switch it on inside a chat.
Two details matter for an owner. First, Claude uses each person's existing permissions in the connected service, so it cannot see anything that person could not already open. Second, on Team plans the owner adds a connector for the organisation, but each person still has to sign in themselves.
In plain terms, MCP connectors are the plugs. You choose which ones go in, and you choose which switches stay off. A sensible first week is one tool, one person and read-only access. Add more only when that works.
What a custom connector is, and when you need one
A custom connector links Claude to a tool through a remote MCP server. According to the custom connectors help page, you enter a name and the server's web address, then choose how it signs in. You can link an existing server or build your own.
You probably need one when:
- your tool is not in the directory, such as industry software or a case management system
- you want Claude to reach your own database or an internal system
- you want a connector built around one specific job, with only the actions you want
Many firms do not need one yet. Start with the ready-made ones. If you do need a custom one, see our connector builds.
Know the catch. The help page says Anthropic has not verified these services, and once connected, Claude can read from and act within them. The connection also comes from Anthropic's cloud, not your device. A server behind a firewall has to allow Anthropic's published IP addresses.
What to check before you connect a tool
This is the part most people skip. Go through the list below first. The checks come from the help pages above, plus plain common sense.
- Permissions. Review what the sign-in asks for. Deny what the job does not need. The help page advises limiting scopes where possible.
- Read-only first. Each connector lists its read and write tools. You can set each one to Always allow, Needs approval or Blocked. Start with reading, and block anything that writes or deletes until you trust the set-up.
- Who approves changes. On Team and Enterprise, owners can set tool permissions for everyone, and users cannot override them. Decide who that owner is, and write it down.
- Shared logins. The help page warns that a shared API key or fixed credential gives every user the same access. Use one with limited scope.
- Logs. At the time of writing (November 2026), the pricing page lists audit logs under Enterprise. If you need a record of who did what, check this before you pick a plan.
- Trust and prompt injection. Text in an email or a document can try to steer an AI. The help page tells you to connect only to servers you trust and to watch tool inputs and outputs.
One more point from the same page. Block any tool that can change things in an outside app until you trust the set-up. You can revoke access at any time in Claude's settings or in the other service.
This guide is general, so check your own contracts and data duties before connecting client systems. If client data or personal data is involved, read your data protection duties first, and see our AI policy template.
Five everyday questions once your tools are connected
These are examples of the kind of thing an owner might ask. What Claude can answer depends on which connector you have and what your account can see. Always check the answers against the source.
- "Look in my Drive folder for supplier invoices and list any older than 30 days." (Google Drive or Microsoft 365)
- "Summarise my last five emails from this client and tell me what they are waiting on." (Gmail)
- "What is in my diary next week, and where am I double booked?" (Google Calendar)
- "Which open deals have had no activity for two weeks?" (HubSpot)
- "Which tasks are overdue, and who owns each one?" (Asana)
Every one of these is a read-only question. That is the point. Get comfortable with reading before you let Claude change anything.
If a question comes back wrong, do not assume the connector is broken. It may be that the source data is out of date or filed in the wrong place. Fix the source and ask again.
When a connector is not the answer
A connector gives Claude access. It does not give it judgement about your business. If your files are messy, the answers will be messy too.
For repeat jobs with sign-off, you may want an AI agent that follows your own rules. Our guide to what an AI agent is explains the difference.
Where Agentic Scale fits
Ready-made connectors cover the common tools. For the system only your firm uses, we build a custom connector, read-only first, with a log of every request.
Connectors are the plumbing. The work is done by AI Agents, our AI employees, which read your files and your connected tools and hand you drafts to approve. We can also build the CRM, portal or dashboard they plug into.
An AI Audit tells you which tools to connect first. It is priced at £197, credited against the build if you commit within 30 days.
Sources
What is a Claude connector?
Do I need a custom connector for Claude?
Is it safe to connect Claude to my business tools?
Which Claude plans can use connectors?
What Claude costs a UK business in November 2026, which plan suits your firm, and how a subscription differs from the API.
5 min read · 12 November 2026Claude is an AI assistant from Anthropic. See what it does well for a small firm, where it falls short, and how to start.
5 min read · 12 November 2026Agentic AI plans and carries out jobs, not just answers questions. Here is what that means, where it helps and where to be careful.
5 min read · 12 November 2026Not sure where to start?Book an AI Audit.
We look at how you work today and tell you honestly what AI should and shouldn't do for you. £197, credited in full if you build within 30 days.