Oracle Analytics agent MCP tools are currently in Limited Availability (LA). Contact us to request enablement for your Oracle Analytics Cloud instance.
Bring Oracle Analytics agents into your AI chat, such as ChatGPT or Claude, or into an agentic workflow.
Through MCP (Model Context Protocol), your AI application can find and invoke the best agent for a given question and return its answer. This allow to interact with multiple Oracle Analytics Agents in the same Chat or AI Workflow.
This walkthrough uses a direct remote MCP connection with OAuth. You need an Oracle Analytics Cloud environment with this capability enabled, access to certified agents, and an AI client that supports remote MCP and OAuth.
1 - Copy your Oracle Analytics connection details
Copy the Client ID: Open your profile from the top-right corner. Under Advanced, turn on Enable developer options and save. Open MCP and copy the Client ID. There is one unique Client ID by Instance.
Profile > MCP, Copy the Client ID for MCP OAuth.
Copy the MCP URL: Open the folder's or agent's Actions menu in Oracle Analytics, select the MCP URL option, and click Copy MCP URL. A folder covers its agents and subfolders; an agent URL targets one agent. For all accessible agents in the instance, copy the agents-only MCP URL from your profile.
2 - Create the MCP Connection in your AI Application
In your AI application, start adding a custom MCP connection, connector, or plugin. Enter the MCP URL (copied from the previous step), select OAuth authentication, and supply the Client ID (copied from the previous step).
Copy the OAuth redirect URL supplied by the AI client that you will need to register in the Oracle Analytics Console (Step 3).
3 - Register your AI client's redirect URL
In Oracle Analytics, open Console > Safe Domains > MCP OAuth Redirect URLs. Add the complete OAuth redirect URL (copied from the previous step). Ask your Oracle Analytics administrator to do this if you cannot edit the setting.
Console > Safe Domains > MCP OAuth Redirect URLs. The screenshot uses an illustrative <id>; enter your complete callback.
4 - Finalise the MCP Connection, sign in and ask a question
Once you have registered the OAuth redirect URL in Oracle Analytics you can finalise the MCP Connection in your AI application. Finish connecting, and complete the Oracle sign-in flow.
Enable the connector or plugin in your conversation if required, then try a question your agents can answer. Your client can discover and invoke agents within the connection's scope.
Example: ChatGPT
In ChaGPT click on Plugin then Add custom MCP server:
- In Connection, select
Server URL and paste your OAC MCP URL. - Set Authentication to
OAuth. - Open
Advanced OAuth settings and set Registration method to User-Defined OAuth Client. - Paste your OAC Client ID into
OAuth client ID. - Copy the Callback URL to register in the OAC console as described in step 3.
ChatGPT custom MCP setup. Instance-specific values are blurred; use your own complete values.
AI client's redirect URL reference
Use the complete redirect URL supplied by your AI client. The references below help you recognize the expected format.
AI client | Redirect URL |
|---|
ChatGPT | Copy the value displayed for your connection. It can use https://chatgpt.com/connector/oauth/<id> or https://chatgpt.com/connector_platform_oauth_redirect. |
Claude hosted apps | https://claude.ai/api/mcp/auth_callback |
Mistral hosted connector | Reference example: https://callback.mistral.ai/v1/integrations_auth/oauth2_callback Confirm the value in your connector setup. |
Gemini / Google integration | Google Account Linking format: https://oauth-redirect.googleusercontent.com/r/<id> Use it only if your selected Gemini integration supplies this callback. |
Always use your client's exact value. A URL containing <id> is a pattern, not a value to paste. Gemini callbacks depend on the integration; the Google Account Linking format is not a universal Gemini MCP callback.
References: ChatGPT authentication, Claude callbacks, Mistral MCP connectors, and Google Account Linking.