MCP interface: Connecting AI assistants
Updated Reading time: 10 min
On this page
- What is MCP?
- The MCP Server in Administration
- Who Can Connect AI Assistants
- Enable Permission for Roles
- Completely Disable MCP
- Security: Which Assistants Are Allowed to Connect
- Allow Only Known Clients
- New AI Assistants: Notify or Approve
- How Approval Works
- Connecting Assistants (Instructions for Employees)
- Your Assistants in Your Profile
- Setup per Program
- Creating Tasks via an AI Assistant
- Logs and Revocation
- Extending Kai with External Tools
- Frequently Asked Questions
Employees connect external AI assistants such as Claude, ChatGPT, or Copilot to your platform via the MCP interface. The assistant always acts on behalf of the person and with exactly their permissions. Conversely, Kai, the platform's AI assistant, can use tools from external services.
Info
The MCP Server page is available to system admins. You control who can connect AI assistants via the "Connect AI assistants via MCP" permission in roles. By default, only system admins have this permission.
What is MCP?#
The Model Context Protocol (MCP) is an open standard that allows AI assistants to access other systems. Kibi Connect uses it in two ways:
External assistants on your platform: Claude, ChatGPT, Copilot, and other programs read content and act on behalf of a logged-in user—for example, searching wiki pages, creating tasks in a group, or scheduling appointments.
Kai with external tools: Kai uses tools from external MCP servers, such as your ERP or CRM system, n8n, or Notion, without employees having to leave the platform.
The MCP Server in Administration#
Open Administration → Integrations → MCP Server. The page has four tabs.
Overview: Under Who can connect AI assistants?, you can see which roles have the permission. Below that, you’ll find the connection URL to share with your employees and a brief explanation of the login process.
Connected Clients: All AI assistants that access your platform, including their usage and last access.
Security: Which clients are allowed to connect, what happens with new AI assistants, and who gets notified.
Instructions & Technical Details: A ready-to-copy text (Copy instructions) for employees to paste into their AI chat. The assistant then guides them step-by-step through the setup. Below that, under Technical Addresses, you’ll find the endpoints in case a program asks for them.
On the MCP Server card under Integrations → Overview, you can also see for whom MCP is enabled: Active for: and the roles, or Off with the note Not enabled for any role.
Warning
Which tools an assistant can use is not determined at the MCP server level. It depends on three things: whether a person’s role has the "Connect AI assistants via MCP" permission, which modules are active on your platform, and what additional permissions the person has. An external assistant can never see or do more than the person themselves is allowed to do on the platform.
Who Can Connect AI Assistants#
Whether a person can connect AI assistants depends solely on their role.
Enable Permission for Roles#
Open Administration → Integrations → MCP Server, Overview tab.
Click Open Roles. The permissions catalog opens directly with the Connect AI assistants via MCP permission.
Check the roles that should receive the permission. Grant to All and Revoke from All change all roles at once.
Click Save.
The Confirm Change dialog shows how many people will gain or lose the permission. Confirm with Save.
For custom and adjustable roles, you can also find the permission in the role editor: Administration → Roles & Permissions, open the role, Permissions tab, System section, AI Assistants (MCP) area. The System Admin role is read-only in the role editor. For this role, you can only change the permission via the dialog above.
Guests: Guests can never connect AI assistants, even via a role.
Without the permission: The person will not see a AI Assistants (MCP) section in their profile. If they still attempt to connect, the consent page will display: "Connecting AI assistants is not enabled for your account. Please contact your administration." No request for approval is sent to the administration.
Permission revoked: If a person loses the permission, the platform immediately blocks their existing connections. If they regain the permission, the connections will work again.
Completely Disable MCP#
To completely disable the interface, click Revoke from All in the dialog above and save. This exception applies to the Connect AI assistants via MCP permission, even for the System Admin role. If no role has the permission, no AI assistant can find the interface: login and registration will respond with "not found" (404), and connected assistants will lose access. In the Overview tab, it will then display Not enabled for any role. There is no separate on/off switch.
Warning
"Reset to Default" restores the permission for the System Admin role. The dialog points this out: If MCP is currently not enabled for any role, it will be enabled again afterward.
Security: Which Assistants Are Allowed to Connect#
Who can connect is controlled by roles. Under Security, you define what is allowed: Any AI assistant with an HTTPS address can register itself with your platform (Dynamic Client Registration). However, it only gains access once a person with the permission logs in and explicitly consents on the consent page.
Assistants that identify themselves via a client metadata document, such as Claude, ChatGPT, and VS Code, appear as Verified on both the consent page and under Connected Clients.
Changes on this tab only take effect once you click Save in the top-right corner of the page. The button will then display Saved again.
Allow Only Known Clients#
With the Allow only known clients toggle in the Allowed Clients card, you restrict new connections to a list of domains. The list is pre-populated with known clients: Claude, ChatGPT, Cursor, Langdock, and Microsoft Copilot Studio. You can remove entries or add your own domains, up to 20 entries (only the address with https://, no path, e.g., https://app.example.com). Use Restore Default List to revert to the pre-populated list.
Assistants on your own computer: Programs like Claude Code that run on the user’s computer remain allowed even with the restriction enabled.
Existing connections: Enabling the restriction does not revoke anything. Therefore, check under Connected Clients if there is anything not on the list and revoke it if necessary.
New AI Assistants: Notify or Approve#
In the New AI Assistants card, you define what happens when someone connects an AI assistant for the first time:
Off: No notification. Each person confirms the connection themselves.
Notify only: The administration receives a notification as soon as someone connects an AI assistant for the first time. This way, unwanted access can be quickly detected and revoked.
Admin approval required: A new AI assistant only gains access once the administration approves it. The person waits until then. Assistants on the Allowed Clients list are already approved. As long as Allow only known clients is enabled, this option is not available, as only what is on the list can connect anyway.
With Notify only, select below under Who gets notified? For approval, select under Who gets notified and approves?: All Admins or Specific Admins. If a selected person is unavailable, notifications go back to all.
How Approval Works#
A person connects an AI assistant whose address is not yet approved. After logging in, they see a waiting page: "Keep this window open." As soon as approval is granted, it automatically continues from there.
The selected admins receive a push notification and a card in the chat. Clicking on it opens the Review AI Assistant dialog: name, address, whether the assistant is verified or unknown, how many people are already connected, and who made the request. Clicking on the name opens a chat with that person.
With Approve, the address is approved: all future connections from the same assistant go through without a new request. With Deny, you can optionally provide a reason, which the person will see in their notification.
If the person closed the waiting page, they receive a notification that the assistant was approved and they can now reconnect it, or that it was not approved, including the reason.
Approving previously denied requests: Under Security → Requests, you can find open and denied requests. A denied request can be approved retroactively with Approve Anyway.
Assistants on your own computer: Local assistants like Claude Code or programs in the development environment (IDE) do not require approval.
Existing connections: If you enable approval, existing connections remain valid.
Connecting Assistants (Instructions for Employees)#
There is no API key or password to share. Each person connects their own assistant.
Info
Prerequisite: One of your roles has the "Connect AI assistants via MCP" permission. If you lack this, you won’t see a AI Assistants (MCP) section in your profile. In that case, contact your administration.
Enter the connection URL in the AI assistant (see setup for each program). You can find it in your profile under the Integrations tab, AI Assistants (MCP).
The assistant automatically logs in to your platform and opens a browser window with the login page.
Log in. On the consent page, you’ll see what the assistant can read and do on your behalf and where the access goes.
Click Allow access.
Done: The assistant now works on your behalf. Access is valid for 365 days.
Tip
If you land on the dashboard instead of the consent page after logging in, simply restart the connection in the AI assistant. Since you’re now logged in, the consent page will appear directly.
Warning
Yellow notice on the consent page: If the application is not a known client, the notice shows the address to which the approval code is sent. Only allow access if you set up the connection yourself.
If the administration has enabled approval for new AI assistants and the assistant is not yet approved, a waiting page appears after login. Keep the window open; once approved, it will continue automatically. If you close it, you’ll receive a notification and can simply reconnect the assistant after approval.
Your Assistants in Your Profile#
In your profile, under the Integrations tab, there is a AI Assistants (MCP) section. It is only visible if one of your roles has the permission.
Connection URL: The address to enter in the AI assistant, with a Copy button, and a brief explanation under How it works.
Your AI Assistants: Your own connections with name, address, label (e.g., Unknown Client), Connected since, Last access, and Active accesses. If admin approval is enabled, your own requests are also listed here: Waiting for approval, Not approved (with reason, if provided), or Approved – reconnect now.
Disconnect: Immediately ends the assistant’s access to your content. This only affects your own connection: other people using the same assistant remain connected. To reconnect, log in again in the assistant.
Instructions for your AI assistant: A collapsible text to copy. Paste it into the AI chat, and the assistant will guide you step-by-step through the setup.
Setup per Program#
Replace your-platform with your platform’s address. The exact connection URL is in the Overview tab and in your profile.
Claude (claude.ai or Claude Desktop): Settings → Connectors → Add Custom Connector. URL:
https://your-platform.kibi.de/api/mcp. Then select "Connect" and log in via the browser.Claude Code: In the terminal, run
claude mcp add --transport http kibi https://your-platform.kibi.de/api/mcp, then in Claude Code, call/mcpand start the login.Langdock: In the Langdock workspace, under Integrations, add your own MCP server, URL
https://your-platform.kibi.de/api/mcp, select authentication "OAuth 2.0 (Dynamic Client Registration)", then connect and log in via the browser.Microsoft Copilot Studio: In the agent, under Tools, add a new tool of type "Model Context Protocol", enter the server URL, select authentication "OAuth 2.0" with dynamic discovery, then create the connection and log in via the browser.
ChatGPT and other programs: Create an MCP server of type "Streamable HTTP" with the connection URL and OAuth login.
Creating Tasks via an AI Assistant#
A connected assistant can create tasks from a meeting protocol for you, for example, with this request: "Create tasks from this protocol in the Sales group, with the keyword Protocol, and attach the protocol to the first task."
In doing so, the assistant can:
Create in a group: The group name is sufficient. This works in your own groups where tasks are enabled. A group task remains unassigned if you don’t specify anyone. Without a group, a personal task is created for you.
Create as a subtask: The task is then attached under an existing task and belongs to its group.
Set and search by keywords: The assistant assigns keywords and finds tasks by keyword again.
Set the status: A new task starts as Todo, In Progress, or Done.
Attach files: The assistant attaches files to a task that you are allowed to see, or a file that it has just uploaded. The task receives a copy; the original remains where it is.
If the assistant cannot find a suitable group of yours, it will not create anything and will report this.
Logs and Revocation#
Logs: Every call from an external assistant is logged.
Connected Clients: The tab shows for each client the name and address, whether it is Verified or uses an Unknown Domain, the active accesses, and the last access. Clicking on a row opens the details with status, Redirect URIs, Used by, and Connected since.
Revoke: Immediately ends a client’s access; all its accesses become invalid. The entry remains in the list and then shows No active access. The assistant must be set up again. Without revocation, access is valid for 365 days.
Disconnect in profile: Each person can disconnect their own connection themselves. Other people using the same assistant remain connected.
Extending Kai with External Tools#
You connect external MCP servers as third-party systems. Detailed instructions are available on the page How do I connect a third-party system to Kibi?.
Open Administration → Integrations → Third-Party Systems.
Click Create Third-Party System and select MCP Server (Tools for Kai) as the Type.
Enter the name, address (HTTPS only), and login details, then create the third-party system.
On the detail page, check which tools the server offers and define who can use them.
With Allow Personal Connectors, employees can additionally connect their own MCP connectors to Kai for their personal use. You can find the toggle in the administration under AI & Workflows in Kai’s Settings. If enabled, employees will find the Connectors tab in Kai’s settings and can set up their connections there. Such connections are only visible to the respective person.
Warning
Data privacy: Requests to external MCP servers leave your platform. Only connect services that you and your employees trust. Every tool usage is logged, and it is visible in the chat when Kai uses an external tool.
Frequently Asked Questions#
Why can’t I connect an AI assistant? Probably none of your roles have the "Connect AI assistants via MCP" permission. In that case, the consent page will show a notice that the connection is not enabled for your account, and the AI Assistants (MCP) section will be missing from your profile. The administration can grant the permission to additional roles.
What data does an external AI assistant see? Only what the logged-in person is allowed to see themselves, and only from modules that are active on your platform.
Can I block individual tools for external assistants? No. You control access via the users’ permissions and the active modules, or by removing the permission from all roles.
What does the yellow warning on the consent page mean? The application is not a known client. Only allow access if you set up the connection yourself. If you want to exclude this entirely, enable Allow only known clients under Security.
Can I review new AI assistants beforehand? Yes. Under MCP Server → Security, set the option Admin approval required for New AI Assistants.
How do I revoke an assistant’s access? In the Connected Clients tab, click Revoke for the client. Each person can also disconnect their own assistant in their profile using Disconnect.
Can an assistant create tasks for a group? Yes, in your own groups, even as a subtask, with keywords, status, and files. More details under "Creating Tasks via an AI Assistant".
What does the MCP interface cost? It is part of Kibi Connect. External services you connect may have their own costs.
How do I completely disable MCP? Remove the permission from all roles using Open Roles with Revoke from All. After that, no AI assistant will find the interface; connected assistants will immediately lose access.