Skip to main content

Govern Gemini Enterprise Agents with JumpCloud AI Gateway

Connect Gemini Enterprise to JumpCloud AI Gateway to control access to enterprise application tools. Admins configure the gateway as a custom Model Context Protocol (MCP) server and select the actions available to Gemini. Users authorize access to their applications in the Gemini Enterprise web app. You can review tool calls in the gateway’s Activity Log.

Prerequisites:

  • Access to the Google Cloud project where you manage Gemini Enterprise data stores and apps.
  • A JumpCloud administrator role that can manage the AI Gateway.
  • At least one MCP server configured in the AI Gateway. See Configure AI Gateway Integrations.

Considerations:

  • The examples use the US datacenter. Use the AI Gateway MCP URL for your datacenter region:
    • US: https://ai.jumpcloud.com/mcp
    • EU: https://ai.eu.jumpcloud.com/mcp
    • IN: https://ai.in.jumpcloud.com/mcp
  • Use OAuth URLs for the same region as your MCP server URL. See the authorization server metadata for US, EU, or IN.
  • This article covers Gemini Enterprise web apps. To connect Gemini CLI, Google’s command-line AI client, see Users: Connect Gemini CLI to the AI Gateway.

Admins: Configuring Gemini Enterprise​

Registering an OAuth Client​

Register an OAuth public client to get the Client ID for Gemini Enterprise. This configuration uses Proof Key for Code Exchange (PKCE) without a client secret.

  1. Run the following command in a terminal:
curl -sS -X POST 'https://ai.jumpcloud.com/oauth/register' \
-H 'Content-Type: application/json' \
-d '{
"client_name": "Gemini Enterprise Client",
"redirect_uris": ["https://vertexaisearch.cloud.google.com/oauth-redirect"],
"grant_types": ["authorization_code", "refresh_token"],
"response_types": ["code"],
"token_endpoint_auth_method": "none",
"scope": "mcp:connect"
}'
  1. Copy the client_id from the response for use in Creating the Gemini Enterprise Data Store.

Authorizing Gemini Enterprise in JumpCloud​

Add Gemini Enterprise's redirect URL to the AI Gateway's trusted list so users can return to Gemini after signing in.

  1. Log in to the JumpCloud Admin Portal.
Important

If your data is stored outside of the US, check which login URL you should be using depending on your region. If your organization uses LDAP, RADIUS, or requires firewall allow list configuration, the Fully Qualified Domain Names (FQDNs) will also be region specific. See JumpCloud Data Centers for the URLs, FQDNs, and IP addresses.

  1. Go to Access > AI Gateway.
  2. Select the Authorized Clients tab.
  3. Click + Add client.
  4. Enter a Title, for example Gemini Enterprise.
  5. (Optional) Enter a description.
  6. In Redirect URL, enter https://vertexaisearch.cloud.google.com/oauth-redirect.
  7. Click Add client.

Creating the Gemini Enterprise Data Store​

  1. In the Google Cloud console, open Gemini Enterprise for your project.
  2. Go to Data stores.
  3. Click Create data store.
  4. For Source, scroll down to MCP Servers > Custom MCP Server, then click Add MCP server.
  5. Under MCP Server Configuration, select OAuth 2.0.
  6. Enter the following values. The URL examples use the US datacenter; use the corresponding URLs for your region:
    • MCP Server URL: https://ai.jumpcloud.com/mcp
    • Authorization URL: https://ai.jumpcloud.com/oauth/authorize
    • Authorization URL Parameters: Leave blank.
    • Token URL: https://ai.jumpcloud.com/oauth/token
    • Client ID: The client_id returned when you registered the OAuth client.
    • Client Secret: Leave blank.
    • Scopes: mcp:connect
  7. Select the Enable PKCE Support checkbox.
  8. Clear the Use HTTP Basic Authentication checkbox.
  9. Click Verify Auth.
  10. Complete the JumpCloud sign-in and authorize the enterprise applications you want to connect.
  11. When Successfully logged in appears, click Continue to finish the authentication settings.
  12. Click Continue to proceed.
  13. Under Advanced options, enter an MCP Server Description that explains what the gateway provides and names the MCP servers you authorized. This helps Gemini Enterprise understand the server's purpose.
    • For example: 'JumpCloud AI Gateway MCP Server provides access to workplace productivity apps, including Atlassian, Slack, and Salesforce.' Include only applications you authorized.
  14. Click Continue.
  15. Under Configure your data connector, select the Location of your data connector.
  16. Enter Your data connector name, for example JumpCloud AI Gateway.
  17. Click Create.

Loading and Enabling Actions​

  1. In the Google Cloud console, open the new Gemini Enterprise data store and select the Actions tab.
  2. Click Reload custom actions.
  3. If prompted, complete tool authorization in JumpCloud.
  4. Review the full list of available actions and select only those you want Gemini Enterprise to access. To expose all available MCP tools, select all actions.
  5. Click Enable actions.

Connecting the Data Store to Your App​

Connect the data store to the Gemini Enterprise app your users access. See Google's Connect a data store to app and authorize Gemini Enterprise for the app connection steps.

To open the app for user testing, select Overview and click Go to Gemini Enterprise.

Users: Authorizing Enterprise Applications​

Authorize the connector with your JumpCloud account and select the applications Gemini can access.

  1. Open the Gemini Enterprise web app provided by your administrator.
  2. Open the connector picker below the chat input.
  3. Click Authorize next to the JumpCloud AI Gateway connector.

Authenticating Your Tools​

  1. Sign in to the JumpCloud User Portal when prompted.
  2. On the AI Gateway authorization page, select the applications you want Gemini to access.
  3. Authenticate to each selected application:
    • OAuth: Click Connect via OAuth and sign in to the application.
    • API Key: Enter your personal API key and click Save. Use the documentation link below the field for help obtaining a key.
  4. When the selected applications show Connected, click Authorize tools.
  5. Return to Gemini Enterprise.

Refreshing Cached Tools​

Gemini Enterprise caches the MCP tool list. After adding an MCP server or adding or removing tools in the AI Gateway, refresh the connector so Gemini Enterprise reflects the changes.

For administrators:

  1. In the Google Cloud console, open the existing JumpCloud AI Gateway data connector and reopen its authentication settings.
  2. Enter the Client ID you used when creating the connector.
  3. Complete the authentication flow: sign in to JumpCloud and authorize all MCP servers you want available in Gemini Enterprise, including newly added servers.
  4. Save the connector changes.
  5. Repeat Loading and Enabling Actions to reload and enable the updated actions.

For users already connected to Gemini Enterprise:

Sign out, refresh the Gemini Enterprise page, and sign in to the JumpCloud AI Gateway again. Complete tool authorization to access the updated tools.

Verifying Tool Calls​

AI Gateway tool calls appear in the gateway's Activity Log and Directory Insights.

  1. Submit a prompt in Gemini Enterprise that uses an authorized application.
  2. In the JumpCloud Admin Portal, go to Access > AI Gateway > Activity Log.
  3. Confirm a tool call event appears for the requesting user.

Troubleshooting​

IssueResolution
Authentication fails, or JumpCloud shows invalid_redirect_uriConfirm https://vertexaisearch.cloud.google.com/oauth-redirect matches both the registered OAuth client and the custom client on the AI Gateway Authorized Clients tab.
OAuth authentication failsCheck the Client ID, blank client secret, and PKCE settings in Creating the Gemini Enterprise Data Store.
Tools are missing or disabled in the data storeFollow Loading and Enabling Actions to reload the tools, complete authorization, and enable the required actions.
Newly added MCP servers or tool changes aren't reflected in Gemini EnterpriseGemini Enterprise caches the configured tools. Follow Refreshing Cached Tools to update the connector's authorization and actions, then have users refresh their sessions.
Users can't see the connectorConfirm the data store is connected to the Gemini Enterprise app they use.
An enterprise app is missing from the authorization pageThe app's MCP server isn't configured in the AI Gateway, or the user doesn't have access to it. See Configure AI Gateway Integrations.

Was this information helpful?