Fueling your DataCamp Classroom with DataCamp's MCP

Summary

A teacher managing a hundred-student classroom can now create assignments, check completion status, and pull a progress report by typing a request into Claude instead of clicking through a dashboard.

That was the case DataCamp marketing manager Mafer made in a webinar closing out the company's "Back to School with AI" series, walking through the newly released DataCamp MCP server in beta. MCP, short for model context protocol, is an open standard that connects an AI assistant like Claude, ChatGPT, or Gemini directly to an external platform, in this case DataCamp Classrooms. Mafer explained the underlying concept, then ran a live demo in Claude: checking her own XP total, getting course recommendations matched to a stated role, listing classroom members, building a new team, creating a graded assignment with a due date, and exporting a CSV of student progress, all through plain-language prompts rather than the DataCamp interface. She closed by demonstrating a separate, already-free feature called Ask AI, a chat panel built directly into the Classrooms dashboard that handles similar requests without needing a Claude or ChatGPT subscription at all. The session also covered a real setup snag: some users on organization-managed Claude accounts won't see the DataCamp connector until their IT team clears it, a limitation Mafer addressed directly rather than glossing over.

Key Takeaways

• Model context protocol (MCP) is an open standard for connecting an AI assistant to an external system, functioning as a universal adapter between tools like Claude and a platform such as DataCamp.
• DataCamp's MCP server is in beta and currently built for Claude; a ChatGPT version is planned but not yet released.
• Both students and teachers can use natural-language prompts through the connector: students can ask for course recommendations or a chart of their own learning activity, while teachers can create assignments, build teams, and pull progress reports for an entire classroom.
• If the DataCamp connector doesn't appear in a Claude account's connector list, the likely cause is an organization-level IT restriction rather than a setup mistake; DataCamp's documentation site addresses common fixes.
• Token usage through the MCP runs on a user's own Claude or ChatGPT subscription, while DataCamp's separate, built-in "Ask AI" feature handles similar requests at no added cost, since it doesn't route through an external AI subscription.
• The tradeoff between the two: the MCP route lets a user combine DataCamp data with other connected tools (visualization, spreadsheets, cloud storage), while Ask AI is self-contained and simpler to get running.

Deep Dives

What Model Context Protocol Actually Connects

Mafer opened with a plain-language definition before touching the demo. "MCP stands for model context protocol," she said. "It's an open source standard for connecting AI applications to any external system." Her framing for the audience: "Think of it as a universal adapter that you can use to connect AI to any tool of your choice, any tool that has an MCP connector."

Applied to DataCamp specifically, that means a user's AI assistant of choice, whether ChatGPT, Claude, or Gemini, can reach directly into a DataCamp account or classroom rather than requiring a separate login and a set of manual clicks through the platform's own interface. Because the protocol is standardized rather than DataCamp-specific, Mafer pitched it as a safer, more consistent way to grant that access than a one-off integration built for a single AI product.

The underlying idea, repeated throughout the session, was that nearly anything a student or teacher can already do inside DataCamp Classrooms becomes reachable through a single written request instead of a multi-step dashboard workflow. Creating a new group of students, checking who has completed an assignment, or pulling a learner's activity history are all existing platform features; MCP changes how a user reaches them, not what exists. "The idea is that you would be able to do pretty much anything that you do either as a learner or as a teacher with just one prompt," Mafer said, framing the shift as one of interface rather than capability.

A Real Setup Snag: Organization Restrictions

Rather than presenting the connector setup as frictionless, Mafer flagged a specific failure mode before it happened to anyone following along live. Connecting the MCP takes a few clicks inside Claude's connector menu, searching for DataCamp and approving access, but she warned that the option sometimes doesn't appear at all. "That may be because you're part of an organization, you're using your organization email, and the IT team at that particular organization has made some sort of restrictions so that you are not able to use the connector," she said.

Her guidance for that case was concrete rather than deflecting the problem: talk to the organization's IT team directly, or check DataCamp's own documentation site, which she gave as mcp-docs.datacamp.com, for workarounds to common access issues. That level of specificity, naming the exact obstacle and where to go next, distinguished the moment from a typical product demo, where such friction is often left out entirely.

The distinction matters because DataCamp Classrooms itself is free for teachers and students through the company's universities program, a point Mafer repeated at both the start and end of the session. A connector blocked by IT policy is a separate problem from the underlying access being paid or restricted, and treating the two as distinct helped set expectations for anyone in the audience running into the same wall after the webinar ended.

What a Student Can Do With a Single Prompt

Mafer ran her first live example from a student's perspective, asking Claude a direct question through the connected MCP: "using DataCamp's MCP, tell me how many XP I have." The answer came back in seconds: 47,000 XP, a number she admitted she had no real benchmark for. "I have no notion of whether that is a lot or not," she said, treating the moment as an honest demo rather than a scripted win.

She then asked Claude to chart her XP by month over the past two years, explicitly requesting a simple visualization rather than a Claude artifact "in the interest of time." The attempt failed on stage: "actually, it tells me that it doesn't have a way to build a chart accurately," she said, and pulled up a chart from an earlier, successful session instead to show what the output looks like when it works. Leaving the failure visible, rather than cutting to a clean take, gave the audience a more honest sense of a beta tool's actual reliability.

Her second student-facing example fared better. Prompting Claude with, "I am a marketing student, I want to become better at AI, recommend me DataCamp courses I should take," produced a specific sequence: an AI Fundamentals track first, followed by an AI for Marketing course and a course on building an AI marketing department, matched to the role she had stated up front. She extended the use case to teachers building a curriculum path for their own students, and noted that a student could just as easily ask Claude to export a list of recommended courses to a spreadsheet for later use elsewhere.

Running a Classroom Through Claude

Mafer switched framing to a teacher managing a live classroom, starting a fresh Claude conversation deliberately. "The more context you give a large language model, the worse the output gets," she said, explaining why she reset before the next set of prompts rather than continuing the same thread.

She asked Claude to check membership in a test classroom she had created for the webinar, then instructed it to take those members and add them to a new team called "AI Challengers." Claude reported no existing team by that name and created one on the spot. "Claude did it on my behalf," she said, pulling up the DataCamp dashboard directly afterward to confirm the team existed and the same members had in fact been added, rather than asking the audience to take the result on faith.

From there she asked Claude to create an assignment, "Introduction to Power BI," due the following Friday, for that same team, and confirmed the assignment appeared correctly in the classroom's assignment list once created. Her final request pushed further into administrative territory: a CSV listing each student's name, email, assigned course, and completion status. She framed the value less around the small classroom used for the demo and more around scale: "imagine that you are a teacher with a hundred people on your class," she said, needing that same status report without manually checking each student's record one at a time. She also noted the same bulk actions extend across multiple classrooms at once, useful for anyone teaching several sections or subjects in parallel.

MCP Versus the Built-In "Ask AI" Panel

After finishing the Claude demo, Mafer pivoted to a second, unrelated way of reaching the same functionality: a chat panel called Ask AI, built directly into the DataCamp Classrooms dashboard rather than routed through an external AI connector. She asked it the same starting question as before, "how many XP do I have," and got the same 47,000 XP answer, showing the two paths converge on identical underlying data.

The distinction she drew mattered most for cost. A viewer question during the session asked directly whether tokens are charged for MCP use, and whether system tokens are limited based on account type. Mafer's answer: usage through Claude runs on a user's own Claude subscription, whether free or paid, so the billing sits with that account rather than with DataCamp. Ask AI, by contrast, doesn't route through an external AI subscription at all, so it carries no separate token cost regardless of plan.

Given the choice, Mafer was direct about her own preference despite the added cost consideration. "I personally prefer to use the MCP in Claude because you can merge it with other tools that are available to you in Claude," she said, pointing to chained actions like turning classroom data into a CSV, feeding it into a separate visualization tool, or uploading results directly to Google Drive, capabilities the more contained Ask AI panel doesn't offer. She demonstrated Ask AI's own range regardless, building a leaderboard of top DataCamp learners by XP and by certifications completed, and creating a new team called "Product Marketing" through the same chat panel, confirming that most core actions work through either path. The choice between them, in her framing, comes down to whether a user needs to combine DataCamp data with other tools or just wants the fastest built-in option with no extra subscription in play.