Configuring the Docebo MCP server (beta)

Introduction

The Model Context Protocol (MCP) server allows you to seamlessly integrate Docebo with enterprise AI assistants (MCP clients) like ChatGPT, Claude, and Microsoft Copilot. Once the integration is configured, users in your organization can interact with Docebo content and data using natural language, directly inside the AI tools they already use, without navigating through the LMS.

Learn more on how to use the Docebo MCP server (opens in a new tab).

This article is intended for Docebo Superadmins and describes how to complete the one-time configuration required to activate the integration.

Requirements and limitations

• A Docebo Superadmin must complete the initial configuration in Docebo.

• The administrator of the target MCP client (for example, ChatGPT Enterprise, Claude, Microsoft Copilot) must have the necessary privileges to add and configure connectors in that platform.

• The current Docebo OAuth flow does not automatically redirect users to the SSO (Single Sign-On) page. Users must already be actively logged into Docebo in their browser before completing the OAuth authentication in the MCP client. If your organization relies strictly on SSO, users may not have direct Docebo credentials and will be unable to complete the flow. This limitation is expected to be resolved in the future General Availability release.

• The current Docebo OAuth flow includes a scope check that could prevent non-superadmins from connecting the Docebo MCP Server to some MCP clients. However, following the step-by-step instructions provided in this article ensures this limitation does not prevent usage with tested clients like ChatGPT, Claude, and Microsoft Copilot. This limitation is expected to be resolved in the future General Availability release.

• If your Docebo platform uses a custom domain, you must use it in place of your original domain when configuring the MCP Server URLs. Please note that while the configuration process for ChatGPT and Microsoft Copilot allows you to successfully set up the integration using a custom domain, the configuration for Claude currently does not. This limitation is expected to be resolved in the future General Availability release.

Configuring the Docebo MCP server

Setting up the integration requires three steps, which need to be completed only once:

1. A Docebo Superadmin creates an OAuth2 application in Docebo.

2. The MCP client administrator connects the AI assistant to Docebo.

3. End users authenticate via OAuth upon their first use.

Step 1: Create an OAuth2 application in Docebo

To allow secure communication between your AI assistant and Docebo, a Superadmin must register the AI assistant as an OAuth2 application within the platform.

Whenever the configuration steps ask you to enter https://[docebo-instance-name], use your platform's domain (e.g., https://yourcompany.docebosaas.com). If your platform uses a custom domain, you must enter your custom domain.

1. Log into your Docebo platform as a Superadmin.

2. Follow the instructions in the Activating and managing the API and SSO app article to create a new OAuth2 application.

3. During setup, use the following configuration:

   • App Name: Docebo MCP Server

   • App Description: App to connect Docebo MCP Server

   • Client ID: docebo_mcp_server

   • Client Secret: automatically generated

   • Redirect URI: the value depends on your MCP client. Use the values listed below for the most common clients. For other clients, refer to their documentation.

     • ChatGPT: the Redirect URI is generated in Step 2. Enter a temporary placeholder for now and return to update this field once ChatGPT provides the URI.

     • Claude: https://claude.ai/api/mcp/auth_callback

     • Microsoft Copilot: the Redirect URI is generated in Step 2. Enter a temporary placeholder for now and return to update this field once Copilot Studio provides the URI.

   • Advanced settings: keep the defaults, with grant types set to Authorization code + implicit grant.

Once the application is created, safely store the Client ID and Client Secret. You will need these values in Step 2.

Step 2: Connect the MCP client

The administrator of your enterprise AI assistant must connect the Docebo MCP server. The steps below describe the setup process for the most common clients. Note that these interfaces change frequently — if the steps no longer match your client's current interface, refer to the client's official documentation: ChatGPT (opens in a new tab), Claude (opens in a new tab), Microsoft Copilot Studio (opens in a new tab).

ChatGPT (Enterprise / Team)

1. Log into ChatGPT with administrative or workspace owner privileges.

2. Navigate to Workspace Settings > Apps.

3. Click Create in the top-right corner.

4. Configure the connection with the following details:

   1. Name: Docebo (You can choose a name that best fits your organization. This name will be visible to final users, so we suggest using the same name they are used to seeing for your Docebo platform.)

   2. MCP Server URL: https://[docebo-instance-name]/mcp

   3. Authentication Type: OAuth

      • OAuth advanced settings:

        • Client ID: paste the Client ID from Step 1

        • Client Secret: paste the Client Secret from Step 1

      • Scopes:

        • Disable the api scope

      • OAuth endpoints:

        • Auth URL: https://[docebo-instance-name]/oauth2/authorize

        • Token URL: https://[docebo-instance-name]/oauth2/token

        • Registration URL: leave this field empty

        • Authorization server base: https://[docebo-instance-name]

        • Resource: https://[docebo-instance-name]

      • Select I understand and want to continue

5. Click Create.

6. ChatGPT will generate a Callback URL. Copy this URL, return to your Docebo OAuth2 app settings from Step 1, paste it into the Redirect URIs field, and save.

7. Enable the connector for the target users.

Claude (Enterprise / Team)

1. Log into Claude with administrative privileges.

2. Navigate to Organizational Settings > Connectors.

3. Click Add custom connector at the bottom of the page.

4. Configure the connection:

   • Name: Docebo (You can choose a name that best fits your organization. This name will be visible to final users, so we suggest using the same name they are used to seeing for your Docebo platform.)

   • Remote MCP Server URL: https://[docebo-instance-name]/mcp

   • Client ID: paste the Client ID from Step 1

   • Client Secret: paste the Client Secret from Step 1

5. Click Add.

6. Enable the connector for the target users.

Microsoft Copilot (via Copilot Studio)

1. Log into Microsoft Copilot Studio as an administrator or environment maker.

2. Select your desired agent, then navigate to the Tools section.

3. Click Add a tool, then select the Model Context Protocol filter tab.

4. Click New tool to add a new MCP server.

5. Provide the following details:

   • Server Name: Docebo (You can choose a name that best fits your organization. This name will be visible to final users, so we suggest using the same name they are used to seeing for your Docebo platform.)

   • Server URL: https://[docebo-instance-name]/mcp

   • Authentication: select OAuth 2.0, type Manual

   • Client ID: paste the Client ID from Step 1

   • Client Secret: paste the Client Secret from Step 1

   • Authorization URL: https://[docebo-instance-name]/oauth2/authorize

   • Token URL template: https://[docebo-instance-name]/oauth2/token

   • Refresh URL: https://[docebo-instance-name]/oauth2/token

   • Scopes: (Leave this field empty)

6. Click Create.

7. Copilot Studio will generate a Redirect URI. Copy this URL, return to your Docebo OAuth2 app settings from Step 1, paste it into the Redirect URIs field, and save.

8. Enable the connector for the target users.

Step 3: End-user authentication

Once the administrator configuration is complete, the integration is ready for end users. Upon first use, users must add the Docebo connector from the list of enterprise connectors made available by their MCP client administrators. After selecting the Docebo connector, they will click the authorization link to initiate the OAuth flow, securely linking their AI assistant profile to their Docebo user account.

Best practice for SSO environments: advise your users to open a new browser tab, log into Docebo via their normal SSO route, and only then click the authorization link inside the MCP client.

Once this one-time authentication is complete, users can access Docebo capabilities directly from their AI assistant going forward.

Available tools

When you interact with Docebo through your AI assistant, it uses specific tools to fetch the right information. Here is a breakdown of the tools available, what they do, and how you can prompt them: