Welcome to the Docebo MCP Server PDG!

Hi! Love this feature. Is there any way to test the MCP in Cursor IDE or other clients (e..g Gemini CLI)? Love the concept but would love to expand the MCP client list a bit if possible. 

Ah, ok, I was able to get it working locally. Thanks.

Hello, We would like to enable the MCP; however, after following the documentation, we noticed that the Docebo MCP Server tile is not available under Artificial Intelligence. Could you please confirm whether this feature is only available in Sandbox environment at this time? Thank you.

Hi ​@Arnaud16619, the updates are currently available in sandbox environments. They will be released to production platforms by the end of this week.

Would be interesting to have an MCP tool for admins that can pull any help article, developer guides, or API docs. Documentation for admins is kind of spread out. I know in the admin UI there’s Harmony which helps - but would be nice to use our own client to steer an agent towards the most up-to-date documentation. 

Hi ​@luca.latini. The Docebo Knowledge Base article for the MCP server gives these examples of questions to ask:

• "What courses have I completed?"
• "Based on my completed courses, what advanced leadership courses should I take next?"
• "I am interested in data privacy. What courses do we have available?"

When I’ve shared this internally, the response has been quite muted. Our concern is that our people are not inclined to ask for or about courses per se, or even be thinking about our LMS at all. They are more inclined to ask questions about specific problems they’re encountering in their work. And opportunities for learning may surface within those conversations.

So, does the MCP server require such… clunky cues to be triggered? Or is this more of an issue with the examples in the help article?

Hi ​@Ian

Thanks for raising this, it's a great question.

The examples in the Knowledge Base are intentionally simple to demonstrate the kinds of information the MCP server can expose, but they shouldn't be interpreted as the ideal user experience.

On the MCP server side, our goal is to provide well-described, generic tools that can work across all Docebo customers, regardless of their industry or internal terminology. Because of that, we intentionally avoid baking in assumptions about specific workflows or prompting patterns.

The experience can become much more natural depending on the AI assistant (e.g. ChatGPT, Claude, etc) you use. AI assistant can add custom instructions or even dedicated agents that use the Docebo MCP server behind the scenes. For example, instead of asking "What courses should I take?", a user could simply ask, "I'm struggling to give constructive feedback to my team," and a well-configured AI agent could recognize that learning content from Docebo is relevant, retrieve the appropriate recommendations through MCP, and incorporate them into the response without the user ever mentioning the LMS or courses explicitly.

So, in short: no, the MCP server does not always require those cues. Those examples are illustrative rather than prescriptive. The quality and naturalness of the experience depends significantly on how the AI assistant is configured to orchestrate and use the available MCP tools.

I'd be very interested to learn more about your setup and the kind of conversations you're trying to enable. If you're open to it, feel free to send me a private message with your email address, and we can arrange a call to go through your use case in more detail and discuss how to make the experience as effective and seamless as possible for your users.