How to Create eLearning with Claude or ChatGPT via MCP

Creating a course with an AI assistant usually means asking for an outline, copying the result into an authoring tool, rebuilding the structure, and repeating the process whenever something changes. Model Context Protocol (MCP) removes much of that handoff. It lets a compatible assistant call approved tools in another product, with your authorization.

For eLearning, that means a request in Claude, ChatGPT, or another MCP client can become a real module in Skillsail: slides, questions, translations, media, version history, publishing, and an LMS-ready export. You still review the instructional quality and final output. The assistant handles the mechanical authoring steps.

The short answer: connect https://skillsail.com/mcp, authorize your Skillsail organization, and ask the assistant to inspect or build a module. It can continue from the initial draft through detailed edits, translation, media, publishing, and SCORM, xAPI, cmi5, or HTML5 export. Review every course before learners see it.

This guide shows the practical workflow, where MCP helps, where it does not, and how the approach differs from using the built-in AI features in a traditional authoring platform.

What MCP changes in course authoring

MCP is not another course format and it does not replace your LMS. It is the connection between an AI assistant and the authoring tools you authorize it to use.

Without MCP, an assistant can draft text for you. With the Skillsail MCP server, it can also:

• Create and inspect modules in your Skillsail workspace.
• Add, edit, delete, and reorder slides, questions, and interactive activities.
• Generate images, video, and voiceover for individual slides.
• Add languages and translate the existing module.
• Set branding and completion rules.
• Publish a module, protect it with a passphrase, or turn it into a reusable template.
• Restore an earlier version.
• Export SCORM 1.2, SCORM 2004, xAPI, cmi5, or standalone HTML5.

The distinction matters: the assistant is not merely describing what a course could look like. It is applying changes to the course in the authorized workspace.

What you need before connecting

You need a Skillsail account and an assistant that supports remote MCP servers over Streamable HTTP. The server address is:

https://skillsail.com/mcp

Client availability varies. Claude, Cursor, and other MCP-compatible tools expose their connector settings differently. ChatGPT supports custom MCP apps, but full write actions currently depend on plan, workspace role, and admin settings. OpenAI's current developer-mode guidance says full MCP support is rolling out in beta for Business, Enterprise, and Edu workspaces on the web. Check your assistant's current documentation before troubleshooting Skillsail itself.

The Skillsail connection guide has the shortest setup path and current troubleshooting notes.

A practical end-to-end workflow

The most useful test is not "can the assistant list tools?" It is whether one conversation can produce a reviewable course and an LMS package.

1. Connect and authorize Skillsail

Add https://skillsail.com/mcp as a remote or Streamable HTTP MCP server in your assistant. The assistant opens a Skillsail sign-in flow. After you sign in, choose the organization the assistant should use.

The connection is organization-scoped. If you belong to several organizations, the assistant only sees the one selected during authorization. Reconnect if you need to switch.

2. Verify the workspace before making changes

Start with a read-only prompt:

Are you connected to Skillsail? Tell me which organization you are using, then list my modules. Do not change anything yet.

This catches the most common setup problem: authorizing the right user but the wrong organization.

3. Ask for a plan, then create the draft

Use a concrete brief and ask the assistant to show its plan before it writes anything:

In Skillsail, plan a five-slide workplace fire-safety module for new office employees. Include evacuation, alarm procedures, extinguisher awareness, and a three-question quiz. Use plain English and aim for five minutes of learning time. Show me the outline first and wait for approval.

After reviewing the outline, approve the build. Skillsail's own authoring system produces the initial module while the external assistant coordinates the request and reports the result.

4. Make precise follow-up changes

Follow-up prompts work best when each one has a clear target:

• "Shorten slide 2 to three bullet points."
• "Add a question about when not to use an extinguisher."
• "Move the evacuation slide before the extinguisher slide."
• "Set score-based completion to 80%."
• "Add German and Spanish."

Open the module in Skillsail after the first substantial change. The assistant can execute the edits, but a human should still check accuracy, tone, accessibility, and whether the assessment measures the intended behavior.

5. Export for the LMS

Once the module is approved, ask for the package your LMS expects:

Export this module as SCORM 2004. Do not change the content. Return the package and summarize the completion settings.

Skillsail can also export SCORM 1.2, xAPI, cmi5, and standalone HTML5. If you are not sure which option to use, confirm the supported standard with your LMS administrator before exporting.

The complete MCP tool reference documents the available authoring, publishing, versioning, media, and export operations.

Security and authorization behavior

A remote MCP server that can edit private content should require authorization. The MCP specification's authorization guidance recommends OAuth for remote HTTP servers that access user-specific data or perform auditable actions.

Skillsail follows that model:

• The assistant acts as the signed-in Skillsail user.
• Access is limited to the organization selected during connection.
• The assistant cannot see other customers' modules.
• The connection does not give Skillsail access to unrelated tools in the assistant.
• Removing the connection in your MCP client stops that client from using the authorized Skillsail tools.

Write actions still deserve care. Deleting a module is permanent, restoring a version overwrites the current module, and changing public visibility takes effect immediately. Ask the assistant to show a plan or request confirmation before destructive or bulk operations.

Practical limitations

MCP makes the workflow faster, but it does not make every assistant or every course equivalent.

• Client support varies. Some assistants only support local MCP servers, some support remote servers but not OAuth, and some restrict write tools by plan or workspace policy.
• The selected organization matters. One connection operates in one Skillsail organization at a time.
• File transfer depends on the client. A client may support tool calls but not pass an uploaded source file to a remote server in the same way another client does.
• AI output still needs review. Verify facts, instructional sequence, quiz answers, translation terminology, alt text, and completion behavior before publishing.
• Complex bespoke interactions may need a specialist tool. MCP is strongest for structured authoring and repeatable operations, not pixel-level animation or arbitrary scripting.
• Exports should be tested in the target LMS. A valid SCORM package can still behave differently because of LMS launch, completion, or reporting settings.

How this compares with Articulate Rise

Articulate publicly positions Rise around block-based responsive authoring and a built-in AI Assistant. Its Q1 2026 release added workflow features such as Magic Import enhancements and AI Course Drafts. In June 2026, Articulate also published an Articulate MCP connection for Claude and ChatGPT.

The current scope is different. Articulate MCP creates a draft outline in Rise and browses existing courses; authors continue building in Rise. Skillsail MCP can create the module and continue with component-level edits, translation, media generation, version restoration, publishing, and LMS export.

That is not automatically better for every team:

• Choose Rise when its polished block system, Review workflow, and wider Articulate ecosystem are the priority.
• Choose Skillsail's MCP workflow when you want the external assistant to continue after the outline, automate repeated changes, translate in the same conversation, generate media, manage versions, publish, or export.
• Pilot both with the same source document when the decision matters. Compare the time to a reviewed, LMS-tested result—not only the quality of the first AI draft.

For the head-to-head decision, read Skillsail vs. Articulate Rise 360. For a wider shortlist, compare the best Articulate Rise alternatives.

A safer prompt pattern for real projects

For anything larger than a single edit, use this sequence:

1. Identify: "Use the module called 'Office Fire Safety'."
2. Inspect: "Summarize its current slides and completion settings."
3. Plan: "Propose the changes, but do not apply them yet."
4. Apply: Approve the exact plan.
5. Verify: Ask the assistant to list what changed, then review the module in Skillsail.
6. Export: Name the required LMS standard explicitly.

This pattern reduces accidental changes and makes it easier to spot when the assistant has interpreted an ambiguous instruction differently from you.

Frequently asked questions

Can Claude create a SCORM course?

Claude can create and edit a Skillsail module through the authorized MCP connection, then ask Skillsail to export it as SCORM 1.2 or SCORM 2004. Skillsail also exports xAPI, cmi5, and standalone HTML5. Test the resulting package in the target LMS before launch.

Can ChatGPT create an eLearning course?

Yes, when the ChatGPT plan, workspace role, admin policy, and app settings allow the Skillsail MCP connection. ChatGPT can direct the same Skillsail authoring operations exposed to other compatible MCP clients. Availability is a client constraint, so check OpenAI's current app guidance.

What is an MCP server for eLearning?

It is an authorized bridge between an AI assistant and an eLearning authoring platform. Instead of returning only text, the assistant can call approved operations that create or change a real course in the user's workspace.

Does Skillsail MCP replace an LMS?

No. Skillsail MCP controls authoring and delivery preparation. Your LMS still enrolls learners, launches the package, tracks completion, and reports results. Skillsail can also publish a direct link when an LMS package is not required.

Can Skillsail MCP translate a course and its voiceover?

It can add supported languages to the module, translate existing content, and generate narration for slides. Translation and voiceover still need human review for terminology, pronunciation, accessibility, and instructional accuracy.

Getting started

Start with a read-only connection check, then build a small disposable module before using MCP on production training.

• Connect your AI assistant to Skillsail
• How Skillsail works in ChatGPT
• How Skillsail works in Claude
• Browse ready-made MCP authoring prompts
• Review every available MCP operation
• Explore Skillsail features
• Compare Skillsail pricing

MCP is most useful when it shortens the distance between a clear instructional request and a reviewable course. The winning workflow is still human-directed: define the audience and outcome, let the assistant handle repeatable authoring work, and verify the result before learners see it.