Create and publish courses with Claude and Codex

Tell your AI what you want to change in your course. The MCP connector links your assistant directly to Teach & Grow: it can read existing content and save your requests in your academy. MCP is the shared connection standard.

• Read courses, chapters and existing lessons for context.
• Create new courses and chapters with titles and descriptions as drafts.
• Create draft text lessons with formatting, source links and practical exercises.
• Insert screenshots and images in lessons, add downloadable course materials and upload videos.
• Edit titles, descriptions, course images, prices, free access and release schedules.
• Reorder chapters and lessons, move lessons within a course and edit exercises.
• Replace or remove images, files and videos from a lesson.
• Check publishing requirements, publish courses with selected chapters and lessons, or return them to draft.
• With additional school permissions: set up the support email, legal notice, privacy policy and existing Digistore24 integration.

Manage a course from first draft to publication through conversation. The same access requirements and plan limits apply as in the course editor. Media uploads require access to the file data and suitable transfer tools. Existing images and embeds are preserved when editing text.

1. Sign in to your academy as a creator. Your assistant can edit existing courses or create a new course with chapters.
2. In your academy, open Settings → Integrations → Claude & Codex. Copy the MCP URL. Always use the address shown there, even if your academy has its own subdomain or domain.
3. Choose your assistant's instructions below. You need your own Claude or Codex account with the ability to add an MCP connection.

1. In Claude, open Connectors under Customize or in settings.
2. Select + → Add custom connector . Enter the name onlinekurshosting and use the copied MCP URL as the server address. Leave extra fields such as client ID and client secret blank.
3. Add the connector and select Connect.
4. On the login page, enter your academy and choose Connect with open academy. If you are already signed in to your academy in the same browser, that session is used. Otherwise, sign in there first.
5. On the permissions page, check the academy name and access to read, edit and publish. Confirm the connection and enable the connector in your Claude conversation if needed.

For team or enterprise accounts, an administrator may need to add the custom connector first. Menu names can vary by Claude version. Official Claude guide.

If you already use Codex CLI, open your terminal and run these two commands. Replace the example address with the MCP URL from your academy settings.

codex mcp add onlinekurshosting --url https://onlinekurshosting.de/mcp
codex mcp login onlinekurshosting

Complete browser login: select your academy, connect using its session and confirm permissions. Then start Codex and use /mcp to check that the connection is available. You can then enter your first request.

Official Codex MCP guide.

If you already use Claude Code, add the connection in your terminal. Replace the example address with the MCP URL from your academy settings.

claude mcp add --transport http onlinekurshosting https://onlinekurshosting.de/mcp

Start Claude Code, enter /mcp and select login for onlinekurshosting . Connect to your academy in the browser and confirm permissions.

Official Claude Code MCP guide.

Give your assistant the file and specify which lesson it belongs in. It must be able to read the actual file data. A filename or path alone is not enough for the remote connector. Even a visible attachment in Claude in the browser is not automatically available as an uploadable file to its tools.

In Codex CLI or Claude Code, you can provide a local file path. If the assistant can access the file and make HTTP uploads, it can transfer it without the course editor. Other clients depend on their available file and upload tools. Before uploading, the assistant checks whether it can actually read the file.

• Screenshots and images: JPEG, PNG or WebP up to 5 MiB and 4000 × 4000 pixels. New images are checked and inserted at the end of lesson content; existing images can be replaced individually.
• Course materials: Files up to 50 MiB, such as PDFs, Office documents, text, CSV or ZIP. The platform's allowed file types apply. Students download them through the lesson's protected downloads.
• Videos: Your assistant transfers videos directly to the video provider and then checks processing. During replacement, the previous video remains embedded until the new one finishes processing and the replacement is completed.
• Course image: Use JPEG, PNG or WebP within the same image limits. Your assistant can add, replace or remove the cover image.

Images and course materials can be replaced, removed or renamed as downloads. Removal detaches the file; previously stored files remain. A new file appears in the lesson only after transfer and completion succeed. The assistant provides the editor link and actual status. An ongoing video upload or processing job is not yet a fully embedded video.

Access for an individual file lasts 30 minutes and can be renewed. Pending image and file uploads expire after 24 hours, after which you start a new upload. Adding media to an already published lesson also requires your explicit approval for an immediately visible change.

Start with a request that checks the connection without changing anything:

Show my courses on Teach & Grow and summarize the existing lessons in “Time management”.

To start a new course with chapters:

Create a course called “Time management for freelancers” with a short description. Add chapters called “Setting priorities” and “Focused work” with suitable descriptions. Save everything as drafts and give me the editor links.

For a new lesson, specify the course, topic, audience and desired outcome:

Create a new lesson in “Time management” about the Pomodoro technique for freelancers. Read the course context first. Research using your available tools and tell me if none are available. Explain the method with an everyday example, add verified source links and a practical exercise. Save everything as a draft and give me the editor link.

To request a specific revision:

Read the “Pomodoro technique” lesson in “Time management”. Make the introduction shorter and more personal, and add an example of a workday with client meetings. Keep the rest of the content. Save the changes and give me the editor link.

If your assistant can read and transfer files, add media with requests like these:

Add the provided “wochenplan.png” file as a screenshot at the end of “Pomodoro technique”. Use the alt text “Weekly plan with focused work blocks”. First check that you can read the file data, then give me the editor link after the upload completes.

Upload the provided “arbeitsblatt.pdf” file as course material for “Pomodoro technique”. Name the download “My first focus plan”. Keep the lesson saved as a draft.

Create a draft lesson called “Focus exercise” in “Focused work” with a short introduction. Upload the provided “fokusuebung.mp4” video. Check processing and finish embedding it once the video is ready. Then give me the editor link and actual status.

Replace the example names with your courses and lessons. Research depends on your assistant's tools; the connector itself does not search the web. Review the draft and its sources before releasing it to students.

You can develop an existing course: reorder chapters and lessons, move lessons between chapters in the same course, and add, revise or remove exercises. Existing learning progress is preserved during moves and edits.

Read “Time management”. Move “Focus exercise” to the start of “Focused work”. Revise its exercise with clearer instructions and preserve students' existing completions.

If you want to delete an exercise that students have completed, the assistant points out the affected completions. It can remove those completions only on your explicit request. Restoring content does not restore deleted exercises or their completions.

You can also change course descriptions, prices, free access and release settings. Set duration, preview access and release times per lesson. Changing release mode may adjust existing lesson schedules; ask for an explanation of the effects before making the change.

Read the current settings of my “Time management” draft. Make the course free and release content immediately after enrollment. Set the estimated duration of “Focus exercise” to 15 minutes. Show me the saved settings.

New courses, chapters and lessons are initially saved as drafts. Your assistant can check the planned course, then publish it together with exactly the selected chapters and lessons. The check itself changes nothing. Unselected drafts remain unpublished.

Check whether I can publish “Time management” with “Setting priorities”, “Focused work” and all their lessons. Do not publish yet. Show missing information and the intended selection.

The check considers the existing publishing checklist, payment and academy settings, and your plan limits. With appropriate permissions, the assistant can fill in missing information. Provide your actual details and requirements.

Publish “Time management” now with the chapters and lessons just checked. Read the latest state again, check the requirements, then return the actual publishing status and links.

The connection needs publishing permission. The course and selected content are saved together: if validation fails, no subset is published. Existing enrollment and release rules continue to apply.

You can also publish individual chapters or lessons, or return courses, chapters and lessons to draft. A chapter or lesson becomes visible to students only when its course and, where applicable, chapter are also published. Unpublishing does not delete content or learning progress.

For changes to already published content, explicitly tell the AI to edit the live content. Depending on its settings, your assistant may request additional confirmation before writing.

Repeated save requests are protected against duplicate execution. If someone has changed the content in the meantime, the outdated write is rejected. The assistant must read the latest state and apply your request to it.

You can view a lesson's MCP change history and restore an earlier content version as a draft. This resets its title, description and content; it does not fully restore exercises, downloads or videos.

Under Settings → Integrations → Claude & Codex , choose Revoke access to revoke a connection. Content read for your requests is processed by the connected AI service.

As an owner or administrator, you can also manage your academy's support email, legal notice, privacy policy and existing Digistore24 integration through MCP. Legal pages can be published directly in the academy or provided as external URLs. These academy settings take effect immediately.

The connection explicitly needs additional permissions to read and change academy settings (school:read and school:write). Existing course connections do not receive these automatically. Request them in your MCP client and reconnect. Check that school permissions are listed on the authorization page; simply signing in again does not expand access.

Read my academy settings. Set “hilfe@meine-domain.de” as the support email and use the legal notice URL I provided. These details should take effect immediately. Use only the data I provided and show the saved links.

For Digistore24, the assistant can set the SHA passphrase, read the actual connection status and assign an existing product to a course. The connector never returns the passphrase. Create your account and product in Digistore24, and send the connection test from there. The connector shows whether a real test arrived and whether it happened after the most recent configuration change.

Check my Digistore24 integration and show the IPN address and any missing steps. Then assign the product ID I provided to my “Time management” draft. Do not invent a successful connection test.

The assistant cannot see any courses.

Check that the connector is active in the conversation and connected to the correct academy. You need an active role with course editing permission.

Login is not working.

Open your academy in the same browser and sign in. Try “Connect with open academy” again. Alternatively, set a password under Account → Profile and use it on the connection login page.

The connector cannot be found or reached.

Copy the MCP URL again from Settings → Integrations. Do not construct an address yourself using your academy domain. Also check that your AI account allows custom connectors.

Saving or publishing is rejected.

Ask the assistant to explain the specific error. Check connection permissions and the course publishing checklist. For version conflicts, the AI must read the current course, chapter or lesson first.

The AI cannot change legal pages or payment settings.

These tools explicitly require school:read and school:write and an owner or administrator role. Request school permissions in the MCP client and approve them when reconnecting. Standard course permissions are not enough.

A content format is unsupported.

For text, use headings, paragraphs, lists, quotes, code, emphasis and links. HTML, tables and strikethrough are unsupported. Insert images through file upload; a Markdown image link is not enough.

The assistant cannot upload my file.

Check that its tools can read and transfer the actual file data. A Claude browser attachment alone does not guarantee this. If needed, use Codex CLI or Claude Code with file access, or upload in the course editor. The assistant can renew expired upload access; after 24 hours, a pending image or file upload needs a new session.

The video is not visible in the lesson yet.

Ask the assistant to check upload and processing status and finish embedding. During replacement, the previous video remains until the exchange is complete. Ask for the actual status and any provider errors.