INTEGRATIONSCLAUDEConnect Claude to your own systems
Preparing a quote with the customer's details to hand, or answering a project question from your own documents: for that, Claude needs access to your data. Anthropic created MCP, the open standard these integrations run on. Here you'll find the routes available, what Claude is allowed to do and what happens to your data.
ISO 9001 & 27001
- Documents and knowledge goes back and forth with Your other systems
- Email, calendar and meetings goes back and forth with Your other systems
- CRM and sales goes back and forth with Your other systems
- Online store and stock goes back and forth with Your other systems
Built for clients including Tata Steel, M&G Investments, Digistore24, Toppers in Concert, Bar-le-Duc and Loogman
This is the data organisations most often let flow between systems. Which of it matters to you depends on where work is getting stuck today.
Documents and knowledge
Files in SharePoint, OneDrive and Google Drive. Claude reads Word, Excel, PowerPoint and PDF, among other formats, and pulls out what's relevant to your question.
Email, calendar and meetings
Outlook, Gmail and Teams, plus meeting transcripts through the Microsoft 365 connector. Useful for preparing a meeting or listing the action points.
CRM and sales
Deals, contacts and conversation history from HubSpot or Salesforce. Both vendors offer their own integration for Claude, with read and write access.
Online store and stock
Products, stock, orders and sales figures from Shopify, through the connector Shopify built for Claude itself.
Financial records
Invoices, contacts and reports from accounting packages with their own MCP server, such as Moneybird. Start with read-only here.
For each combination: which data moves back and forth, what the standard integration does and what to watch out for.
- Connect Exact Online to ClaudeConnect Exact Online to Claude: your own MCP server as a custom connector, read-only access to your books, and what you need to arrange for the GDPR.
- Connect Claude to HubSpotConnect Claude to HubSpot with HubSpot's own connector: pipeline, customer history, tickets and marketing content. What it can do, and what to set up first.
- Connect Claude to SalesforceConnect Claude to Salesforce through the hosted MCP servers or the Salesforce in Claude plug-in. What Claude can read and change, and what to set up first.
- Connect Claude to Microsoft 365Connect Claude to Outlook, Teams, SharePoint and OneDrive with Anthropic's Microsoft 365 connector. What Claude can do, and what your admin needs to set up.
- Connect Claude to ShopifyConnect Claude to your Shopify store with Shopify's own connector: products, inventory, orders and sales. What Claude can do, what it can't, and what to watch.
Not always. If an existing integration for Claude does exactly what you need, we'll tell you.
Anthropic built its own connector for Microsoft 365 and Google Workspace. HubSpot, Shopify and Salesforce offer their own integration with Claude; for Salesforce, since September 2026 that includes a beta plug-in with ready-made sales workflows. For searching, summarising and simple changes, a connector like that is often enough.
On Team and Enterprise, an owner decides which connectors are available and what they're allowed to do. Enterprise also has audit logs and a Compliance API, and can stop staff from connecting a work account to a Claude account outside the organisation.
A connector falls short when your system doesn't have one, when Claude has to follow rules the connector doesn't know, or when the outcome has to land in a fixed process. We can connect any system, even when there's no standard integration.
- Your system has no connector, for example an in-house database or an industry-specific package.
- Claude needs to carry out actions the standard connector doesn't offer.
- The server sits in a closed network and shouldn't simply be reachable from the internet.
- You want Claude in your own software or customer portal, with your own approval steps.
- Your data has to stay in Europe and you want to use Claude through a European cloud region.
Connectors from the directory
Anthropic built its own integrations for Google Drive, Gmail, Google Calendar, Slack and Microsoft 365, among others. The directory also lists connectors from vendors, such as HubSpot and Shopify. A tick means Anthropic has tested the connector more closely, but it isn't a security audit. On Team and Enterprise, an owner switches a connector on first, then each employee signs in themselves.
Your own connector via remote MCP
For a system without a connector, you build an MCP server and add it as a custom connector. On Team and Enterprise, only an owner can do that. Claude connects from Anthropic's cloud, including in the desktop app. So the server has to be reachable over the internet; if it sits behind a firewall, you add Anthropic's IP addresses to the allowlist.
Through the Claude API
In your own application you use tool use: you define the tools and your code runs them, so you decide where a person has to approve. With the MCP connector (beta), the API connects directly to a remote MCP server and you switch individual tools on or off. That connector isn't covered by zero data retention.
Permission per action
For each connector, you set whether a tool is always allowed, asks for approval first or is blocked. Team and Enterprise owners set this for the whole organisation, for example: reading email is allowed, sending isn't. It never goes beyond the permissions in the source system.
Privacy and data residency
By default, Anthropic doesn't train on data from Team, Enterprise or the API. The data processing addendum, including the EU standard contractual clauses, is automatically part of the commercial terms. Anthropic stores data in the US. If you need storage and processing in Europe, you use Claude through Amazon Bedrock, Google Cloud Vertex AI or Microsoft Foundry.
An integration moves data. In between, AI can read, sort and prepare a suggestion. Where it matters, a person decides.
Research without write access
In Research mode, Claude can call your connectors' tools without asking for approval each time. Anthropic recommends switching off write tools on custom connectors in that case. That way a research question stays a research question.
Sending stays a human click
Claude prepares drafts: a reply in Outlook, an opportunity update, a product description. In the Microsoft 365 connector, sending email and Teams messages can't be set to 'always allow'.
Figures from the system
Have Claude fetch amounts and quantities with a query or report from the source system, not estimate them. Ask for record numbers or links, so you can check.
Permissions that are too broad at the source
Claude inherits the user's permissions. The Microsoft 365 connector, for example, searches all of SharePoint within those permissions; you can't limit it to a single site. If folders are shared too widely, they'll surface in Claude too.
A server you don't know
A malicious MCP server can send hidden instructions. A connector's developer defines the tools and can still change them after a review. Only connect servers from parties you trust, and watch for behaviour that changes.
Network policies that block everything
Connections come from Anthropic's IP range, not from the employee's device. Location-based access rules in Microsoft Entra therefore block the Microsoft 365 connector for everyone. That needs a deliberate exception.
Misjudging data residency
The Claude apps store data in the US, and connected services process data on their own infrastructure. If you have an EU requirement, arrange it up front through a cloud region and check where the connected services run as well.
What goes where
Which data moves, in which direction, and which system wins when two values differ. That is written down before anything gets built. For this integration, that means: Documents and knowledge, Email, calendar and meetings, CRM and sales, Online store and stock, Financial records.
Fields and rules
For every field, how it translates between the systems: statuses, VAT codes, general ledger accounts, item numbers. And what happens with a duplicate or a missing value. For this integration, that means: Connectors from the directory, Your own connector via remote MCP, Through the Claude API, Permission per action, Privacy and data residency.
Testing with real cases
In a sandbox or test administration, with the exceptions you actually run into: credit notes, partial deliveries, changed customer details.
Live, with logging and monitoring
Every sync is logged, and when something fails, someone gets an alert. If a vendor changes its API, we update the integration.
No account manager. You talk to someone who builds integrations themselves, and who will tell you when a standard integration is enough.
- Office on the Duivendrechtsekade in Amsterdam
- ISO 9001 and ISO 27001, certified by Kiwa
- 500 projects delivered
- 4.8 from 127 reviews
Does Anthropic use our data to train Claude?
Not by default. Anthropic doesn't train on inputs and outputs from commercial products such as Claude Team, Enterprise and the API. The exception is feedback: if an employee gives a thumbs up or down, that conversation may be used. Team and Enterprise owners can switch off the feedback button.
Where does Anthropic process our data?
Anthropic stores business customers' data in the US and may route traffic to countries in the US, Europe, Asia and Australia. Through the API you can restrict processing to the US, but not to Europe. For storage and processing within Europe, you use Claude through Amazon Bedrock, Google Cloud Vertex AI or Microsoft Foundry.
Is there a data processing agreement with Anthropic?
Yes. Anthropic's Data Processing Addendum is automatically part of the commercial terms and includes the standard contractual clauses for transfers outside the EU. On Team and Enterprise, your organisation is the data controller and Anthropic is the processor.
What's the difference between a connector and the API?
You use a connector in the Claude app: staff ask questions and Claude fetches data or carries out an action. With the API, you build Claude into your own software, with your own interface, logic and approval steps. Both can use the same remote MCP server.
Do customers need to know they're talking to Claude?
Yes, if you use Claude in contact with people, for example as a chat assistant on your website. Under the AI Act, people must then know they're dealing with AI; those transparency rules have applied since August 2026. For decisions about people, such as recruitment, stricter requirements apply from 2 December 2027.
Can Claude do the wrong thing in our systems?
Only what the connector offers and what the user is allowed to do in the source system. Set write actions to 'needs approval' and block what you don't want, such as deleting. Only choose 'always allow' for tools and servers you trust.
Sources
The technical facts on this page come from the vendors' official documentation, checked on 19 September 2026. Vendors change their APIs and limits; when we build an integration, we always check the current state.
- Claude Help Center: Use connectors to extend Claude's capabilities
- Claude Help Center: Get started with custom connectors using remote MCP
- Claude Docs: Connectors overview
- Claude Docs: Connector verification
- Claude Platform Docs: MCP connector
- Claude Platform Docs: Tool use
- Claude Platform Docs: Data residency
- Anthropic Privacy Center: Is my data used for model training?
- Anthropic Privacy Center: Where are your servers located?
- Anthropic: Data Processing Addendum
- Claude: Regional compliance
- Claude Help Center: Access audit logs
- Claude Help Center: Microsoft 365 connector security guide
- European Commission: AI Act, regulatory framework for AI