By now, we're used to the idea that ChatGPT helps you write texts, summarize documents, and brainstorm new ideas. But what many people don't know yet is that ChatGPT has recently also been able to communicate with your own applications, documents and data sources. This transforms this chatbot from a smart assistant into a fully-fledged one. personal digital assistant.
Until recently, ChatGPT could mainly talk about general, public knowledge. Things that you can also find via Google, such as summaries of reports, explanations of concepts, or writing texts based on universal sources.
The big breakthrough came when OpenAI activated a feature in the spring of 2025 that allowed users to connect ChatGPT to personal data and systems. This includes your calendar, email, cloud storage, CRM, or even your internal databases. Suddenly, ChatGPT was not only able to provide generic answers, but work specifically with your company information.
The technology behind this breakthrough is called it Model Context Protocol (MCP).
The Model Context Protocol was introduced by the AI company in late 2024 Anthropic — known for their AI chatbot Claude. In the most recent versions of Claude, we already saw how the bot was able to connect securely and in a structured way to external systems via MCP.
💡 MCP is now also called “the USB-C port of language models” by experts: a standardized, universal way to connect AI models to other systems, databases and applications. Just like USB-C makes it possible to connect any device — regardless of brand or platform — MCP does so for language models.
MCP thus not only makes generative AI possible, but also introduces a form of interactive, contextual AI: an AI that understands your specific situation and performs tasks independently within it.
OpenAI recently added this new functionality to ChatGPT, in the so-called “Connectors” environment. The option is available to users of the paid versions, including ChatGPT Plus and Team accounts.
Here's how to activate it:
This functionality is remarkably user-friendly: most connections can be set up with a few clicks and work directly from the web environment. That makes ChatGPT accessible to both technicians and non-technical users.
.webp)
It is also possible to work with MCP in Claude, Anthropic's AI assistant. But this requires a little more technical actions. This is because the function is hidden a little deeper in the settings of the desktop version.
Follow these steps:
So while Claude's flexibility and strength is impressive, it requires more technical knowledge to set up an integration. While ChatGPT makes everything accessible via a simple GUI, Claude is especially suitable for developers and advanced users who want maximum control over the connections.
The applications are endless. Here are a few practical examples:
Simply ask ChatGPT, “What appointments do I have today?” or “What is my weekly schedule?” and you get an immediate overview based on your linked Google Calendar or Outlook.
Let ChatGPT summarize your emails, search for specific topics, or ask, “What was the last email from customer X about quote Y?”
For example, with Google Drive or Dropbox. Question: “What are the latest updates to customer Z's project file?” and you get a brief answer from your own documents.
For example, you can link an internal sales database or Airtable and ask, “What is the status of project Delta?” or “Add a new record with this information.”
Think HubSpot or Salesforce. Question: “What leads should I follow up on today?” or “What was the last interaction with lead X?”
A common question about these types of breakthroughs is: Is this already Agentic AI? The short answer: not quite yet — but it's getting awfully close.
The current application of MCP within ChatGPT and Claude is what you might call a 'light variant' of Agentic AI. The AI does not yet have an overview of the entire process itself, but it can:
Where a full-fledged Agentic AI itself orchestrates tasks and decisions — like a project manager in software form — this setup acts more like a smart, always available employee who executes your assignments and understands context.
The big step that comes next? That the AI itself will plan, decide and combine tasks into complete processes, with you as the final director.
But what we see here now is the infrastructure on which it will soon be built.
Curious about which systems can already be linked via MCP? The open source community has compiled an overview of existing servers that you can use or expand immediately. Check out the list here:
👉 https://github.com/wong2/awesome-mcp-servers
At The Automation Group, we linked, among other things, Airtable. This allows us to request, modify and even create records automatically via the chat environment. As a result, our internal workflows have become considerably faster and more consistent.
What makes this development so special is that it fundamentally changes how we use AI. ChatGPT is no longer only used for separate prompts such as text writing or brainstorming, but is growing into a full-fledged colleague that switches independently between tools, retrieves information, analyses and executes actions.
Especially for entrepreneurs, consultants, marketers and managers, this means an enormous acceleration in productivity. And for larger organizations, it is the step towards integrating AI into operational processes.
The introduction of Connectors and MCP is a silent revolution. It may not be a feature that makes you jump right away, but it is one that structurally changes your work.
So: are you already using it?
What would you like to connect ChatGPT or Claude to?
And which repetitive tasks would you like to automate?
Let us know — we're curious about your application. And we'll be sharing more examples and guides on ai.nl soon.
.svg)
