Use MCP integrations to connect Robin and AI Copilot to the tools your team already uses. Once connected, your AI assistants (Robin and AI Copilot) can pull relevant context and perform supported actions - without switching between systems.
MCP (Model Context Protocol) is a standardized way for AI assistants to securely interact with external tools and services. MCP integrations can be enabled for Robin, AI Copilot, or both.
MCP integrations are managed from the AI Center, where you can install out-of-the-box integrations from the MCP Catalog, add and manage custom MCP servers, and control who can use each integration (Robin, AI Copilot, or both).
MCP integrations respect the RBAC (Role-Based Access Control) model of each connected system. Actions and data access are limited by the permissions granted during authorization.
Add MCP server
Use the MCP Catalog to quickly connect Robin, AI Copilot, or both to the tools your team already uses - so they can pull the right context and take action without jumping between systems.
For custom or internal integrations, you can add your own MCP server to connect a private tool, a custom workflow, or anything that isn't in the catalog, while keeping full control over what capabilities you expose.
Add integration from catalog
Note: Some catalog integrations require additional setup in the third-party application.
Examples:
- Monday: A Monday.com account admin must install the Monday MCP app from the Monday Marketplace. After this step is completed, the integration can be connected in Atera and authorized by users.
- Notion: No marketplace installation is required. Admins enable the integration in Atera, and users authorize it on first use.
From the MCP Catalog, you get prebuilt integrations for popular tools like Monday, Box, Notion, Intercom, Asana, and Atlassian. Each catalog integration comes with the MCP server details prefilled (name, URL, and description), so you can install it in a click, choose who can use it in your workspace, and then connect to authorize access. The catalog is added to continuously. Open AI Center > MCP integrations to see what is currently available in your account.
To install a catalogued MCP server:
1. Go to AI Center > MCP integrations.
2. Click Install on your chosen app.
The Add MCP server window appears.
3. Update the name, description, and audience (All, Robin, AI Copilot) if needed.
4. Click Create.
The server is added under Installed MCP servers.
To complete the setup, see Connect installed MCP servers
Add custom integration
For custom integrations, you can add your own MCP server to connect tools that aren't in the catalog, like internal systems, private APIs, or custom workflows. Just enter the server name, MCP endpoint URL, a short description of what it enables, and who should have access in your workspace, then connect to authorize and start using it.
To install a custom MCP server:
1. Go to AI Center > MCP integrations.
2. Click Add MCP server (top right of page).
The Add MCP server window appears.
3. Enter the server details:
- Name: Enter the name.
- Server URL: Enter the MCP endpoint URL (for example, an SSE endpoint).
- Description: Enter a description (optional).
- Audience: Select All, Robin, or AI Copilot.
4. Click Create.
The server is added under Installed MCP servers.
To complete the setup, see Connect installed MCP servers
Hardening a custom MCP server
Covering three things:
- Your endpoint must support OAuth dynamic client registration, or have pre-registered client credentials. Without either, authorization fails at the Connect step.
- Expose the narrowest useful capability set. There's no per-tool control in Atera, so whatever your server implements is what Robin can call. Start read-only; keep destructive operations out of the server rather than trusting Robin not to pick them.
- Set the Audience deliberately, and know that connecting the server makes it available but grants nobody access — each user authorizes separately, bounded by their own permissions.
MCP server or cloud action? — use MCP when Robin should explore a system across many requests; use a cloud action for one specific call. If you're building an MCP server for a single operation, you want a cloud action.
Edit MCP server
Edit an installed MCP server to update its details (name, description, server URL, or who it's available to) so the integration stays accurate as your setup changes.
Note: You can quickly update the audience using the blue dropdown menu.
To edit an installed MCP server:
1. Go to AI Center > MCP integrations.
2. Click the ellipses on the installed MCP server. Then click Edit.
The Edit MCP server window appears.
3. Make any changes. Then click Save.
Delete MCP server
Delete an MCP server when you no longer need the integration or want to remove it from your workspace. This removes the server from your installed list (and disconnects it if it was connected).
To delete an installed MCP server:
1. Go to AI Center > MCP integrations.
2. Click the ellipses on the installed MCP server. Then click Delete.
Connect installed MCP servers
After adding an MCP server, connect it to authorize it with the provider. This final step activates the integration for the selected audience.
To connect to an installed MCP server:
1. Go to AI Center > MCP integrations.
2. Click Connect on the installed MCP server.
3. You'll be redirected to the provider to finish the setup. Follow the on-screen prompts to sign in, review the requested access, and approve the connection. When you're done, return to Atera and the integration will be ready to use.
After an admin installs and connects an MCP integration, each user must authorize it the first time they use it. Authorization is performed via OAuth and is per user, not company-wide. This ensures MCP integrations respect the permission model (RBAC) of the connected system.
Robin usage note: To activate an MCP integration in Robin, end users must ask a direct question that references the connected tool (for example, "Find our onboarding checklist in Notion"). If it's their first time using the integration, they'll see an authorization link and be guided through the approval flow.
Example: connect Notion from start to finish
The sections above are grouped by task rather than in the order you would work through them. Here is the full sequence for a catalog integration, using Notion as the example.
- Go to AI Center > MCP integrations and click Install on Notion. Notion needs no setup in the third-party application first, unlike Monday.
- In the Add MCP server window, set the Audience to All, Robin, or AI Copilot, and adjust the name and description if you want. Click Create.
- Notion now appears under Installed MCP servers, but is not yet usable. Click Connect on it.
- Sign in to Notion, review the access being requested, and approve it. You return to Atera with the integration active for the audience you chose.
- Each user authorizes separately the first time they use it. In Robin, a user triggers this by asking a question that names the tool, for example "Find our onboarding checklist in Notion". They are shown an authorization link and guided through the approval.
Note: Step 5 is per user and cannot be done on their behalf. Installing and connecting the integration makes it available; it does not grant anyone access. Each person's access is limited to what their own permissions in the connected system allow.
