AI Assistant and MCP server for Seatsurfing
"Is there a desk near Anna tomorrow?" Now you can simply ask. Seatsurfing Cloud gets two new AI features: an AI Assistant in the booking UI and an MCP server that connects your organization's own AI tools.
💬 AI Assistant: book by chatting
The new Assistant item in the booking UI opens a chat panel. Your users can:
- find free spaces in plain language,
- book a space and list their bookings,
- cancel a booking,
- see where their buddies are booked and book a seat near them.
For now, English and German are supported. More languages may follow later.
You stay in control
The assistant acts on behalf of the signed-in user and can do exactly what this user can do in the booking UI, and nothing else. All booking rules and permissions apply. Bookings and cancellations are never executed on their own: the assistant shows a card with the details, and only after the user confirms it, the action runs.
Data protection
The assistant uses a language model in Azure AI Foundry, deployed in Germany West Central. Conversations are deleted 24 hours after their last message. The assistant is on by default and organization admins can switch it off on the Cloud Features page. Please note that the names and booked seats of buddies are processed by Azure when a user asks about them.
🔌 MCP server: use your own AI tools
If your organization already works with Claude, ChatGPT, Copilot or other tools supporting the Model Context Protocol (MCP), you can connect them directly to Seatsurfing. Each user signs in with their own account and allows access on a consent page, so the tool can only do what the user could do themselves.
Direct access is off by default. Admins enable it under Cloud Features ("Access for your own AI tools") and find the server URL for their organization there. Switching it off again disconnects all tools right away.
The two switches are independent: the assistant keeps working if direct MCP access is off, and the other way around.
☁️ Available for Seatsurfing Cloud
Both features are Cloud-only. They are not available for self-hosted installations.
Read the documentation for all the details.
