Connect Your AI Assistant

Scormbusters never makes an AI call and never stores a model key. Instead, it is an MCP server: your own AI assistant connects to it and works on your courses through documented tools — reading and writing lessons and blocks, editing Scenes slides and objects, looking at a slide's pictures to describe them, building branching, running accessibility audits, handling review feedback, importing and exporting packages.

That means the AI you use is the one you already pay for (a Claude or ChatGPT subscription, a Claude Code seat, or a model running on your own hardware through any MCP client), the conversation happens in that assistant's own app where you approve every change, and Scormbusters never bills you for AI — there are no credits, no allowance, and nothing that expires.

1. In Scormbusters, open Settings → AI assistant (the page is headed Connect your assistant) and copy your MCP server URL.
2. In Claude, open Customize → Connectors, select + Add, then Add custom connector. Name it Scormbusters, paste the address, and keep the recommended sign-in option.
3. Sign in to Scormbusters when asked and approve the connection. The approval page says what it may do: read or author courses, matching your role; it cannot delete courses, manage members or change billing.
4. Try *"List my courses."*

Claude Code: run claude mcp add --transport http scormbusters <your MCP server URL>, then type /mcp, choose scormbusters and sign in. The settings page shows the command with your address filled in, as well as the commands for Codex CLI and Gemini CLI.

ChatGPT: add Scormbusters as a custom MCP server with the same address and choose OAuth sign-in (your plan or workspace may not allow custom MCP servers).

Cursor and VS Code: the settings page has one-click install buttons.

The Model Context Protocol is an open standard, so any compliant client works; setup for each one is in the MCP reference.

Some apps — open-model front ends such as LM Studio or Open WebUI, scripts, older clients — cannot do the OAuth sign-in. Give them a personal token instead:

1. Open Settings → API, section Personal tokens — connect an AI assistant.
2. Give the token a Name, choose when it Expires in (30 days to 1 year; 90 days by default), and tick Read-only if the assistant should only look.
3. Create token, and copy it at once — it is not shown again.
4. Configure the app to send Authorization: Bearer oc_pat_… to your MCP server URL.

Any member can create personal tokens for themselves. A personal token works with the MCP server only; the REST API takes organization API keys (see API keys and webhooks).

The assistant acts as you: a Viewer's assistant can only read, an Owner's, Admin's or Member's can write content, and nothing it does bypasses your organization's permissions. Changes it makes are recorded in the audit log.

Every OAuth connection is listed in Settings → Connected apps with what it may do and when it was last used; Revoke cuts it off at once. Owners and Admins see every connection in the organization. Personal tokens are revoked from Settings → API.

The full tool reference, with example conversations, is at /docs/mcp.