The MCP server
Everything an AI tool needs to create explainers is exposed as a remote MCP server:
https://scrimba.com/explain/mcpClaude Code and Codex have their own setup pages. Any other MCP-capable client (Cursor, Windsurf, Zed, your own agent) connects the same way: add the URL as a remote streamable HTTP server, the way your tool adds MCP servers.
What the connected tool can do
The server gives the agent tools to open an explainer, stream the lesson to Scrimba slide by slide, and finish it, plus one that groups 2 to 12 explainers into a playlist you claim with a single link.
The agent is the author: it writes the narration, the code excerpts, the diffs, and the diagrams itself, which is what makes explainers of your private code possible. Scrimba renders whatever it streams, live, as it arrives.
Better explainers
The connected model is the author: it writes the lesson and the narration itself, and Scrimba renders what it streams. A stronger model, or one told to work harder on the lesson, produces a better explainer.
Drafts and claiming
An explainer created through this server starts as a private draft. Nobody can stumble onto it, and the link alone isn't enough to watch it.
The creation link your agent hands you is the key: opening it and signing in to Scrimba claims the explainer as yours. From then on it's in your account and you control its visibility. See Privacy, claiming and sharing.
If you want to share it straight away, ask your agent to make the explainer unlisted: once you've claimed it, anyone with the link can watch. Explainers made through ChatGPT and through the agent plugin and CI endpoint work differently and start unlisted.
Creating drafts is free: the connected agent authors the lesson, so no Scrimba AI generation is involved. Images are generated the first time the link is opened, and narration once the explainer is claimed and watched.