What are AI Connections in Task Manager?
AI Connections in Task Manager let you link your account with ChatGPT or Claude. In a conversation with the model, you then ask about your tasks, chats, notes, files, mail and calendar, and the model searches Task Manager for the content it needs and reads it. The connection works through MCP (Model Context Protocol), an open standard that ChatGPT and Claude use to connect with other apps.
The model has read-only access: it cannot add, change, send or delete anything. Each person connects their own account and sees through the model exactly what they see in Task Manager - with the same permissions, nothing more.
AI Connections belong to the Pro plan. The feature is enabled by the administrator for the whole company.
How does it work?
The administrator enables the feature in Settings → Advanced → AI Connections (MCP). If the company has data encryption enabled, the app first shows a warning: encrypted content will be decrypted and sent to the AI provider. The feature only switches on once the consent box is ticked, and the app records who gave it and when.
Then everyone connects their own account. In Settings → Profile → AI Connections (ChatGPT, Claude), they copy their Task Manager address, then in ChatGPT or Claude add a new connector and paste in that address. The login page of your Task Manager opens. You log in with your email and password, choose how long the connection should last - 90 days by default - and click Allow.
From that moment on, whenever the model needs data, it searches for it in Task Manager and retrieves the full content of a task, chat, message, note, file or mail thread. A task arrives together with its description, people, To-do list and list of attachments. The model also checks the calendar for the chosen period - task due dates, your events and reminders, and team absences - so it can answer "what have I got next week?". From the AI Knowledge Base it gets summaries, decisions and conclusions from the tasks, chats, notes and files you have access to.
What do you gain?
You ask your questions right where you already work with AI. You don't paste the content of tasks and emails into ChatGPT - the model finds what it needs in Task Manager itself to answer you.
You're not giving up control. The model only reads, and sees the same as the person who connected it. Anyone can disconnect their own connection at any time, and the administrator sees all connections and can disconnect any of them.
You know what the model has read. The connection log shows who read data, when, with which app and which tool - without showing the content itself. Entries older than 180 days are removed automatically by the app.
You can connect more inboxes without worry. You can set an inbox in Mail as Read-only: Task Manager won't send anything from it, including in Automations, while the AI Assistant and connected models still know that correspondence.
An everyday example
Marek runs a design office and plans his week in a conversation with Claude. On Monday he asks: "What did we agree with the Nowak company about the delivery date?". Claude searches for "Nowak" in Task Manager, reads the task with the arrangements and the latest mail thread with the client, then summarises what was promised and by when.
Marek also asks for a draft reply to the client. Claude writes it in the conversation, but won't send it from Task Manager, because the connection is read-only. Marek pastes the text into the thread and sends it himself. Finally, he asks what he has next week, and gets the task due dates, his own meetings, and who on the team will be on leave. He connected the invoices inbox on his end as read-only, so Claude knows that correspondence too, but Task Manager won't send anything from it.
Frequently asked questions
Are AI Connections included in the Standard variant?
No, that's a Pro variant feature. Prices for both variants are listed in the pricing, in amounts excl. VAT.
Can ChatGPT or Claude change anything in Task Manager?
No. The connection is read-only: the model won't add, change, send or delete anything. Even if someone hides an instruction in the content of an email, the model won't make any change in Task Manager.
What does the model see?
Exactly what the person who connected it sees: tasks with their To-do list and list of attachments, chats and messages, notes, files, mail from their own inboxes, the calendar (task due dates, events, reminders and team absences), and the AI Knowledge Base. It never sees passwords, API keys or settings, and other people's private conversations with the AI Assistant are excluded.
Where does the content that's read end up?
With the company that provides the AI app: OpenAI for ChatGPT and Anthropic for Claude, even outside the European Union. The app's data still stays on your server. If encryption is enabled, the administrator must agree that the content will be sent to the provider decrypted.
How long does the connection last?
90 days by default - you choose the validity period when you give consent. You can leave the field empty and connect without an expiry, but the app advises against it. The connection also ends when disconnected, when the password is changed, and when the account is blocked.
What do I need to make this work?
Task Manager must be accessible from the internet over HTTPS, because ChatGPT and Claude connect from their own servers. On the AI side, you need a ChatGPT or Claude plan that allows adding your own connector. The connection doesn't use the company's API key from the AI Assistant settings.