Write video tutorials with an AI agent

When you write video tutorials with an AI agent, you’re asking Claude Code, Cursor or Antigravity to do what they already do well: read your code and write files. Your agent already knows your screens, your buttons and your flows. With Ondaclip, it turns them into tutorials your team reviews just like code.

Write a video tutorial with an AI agent in a single prompt

In your editor, type: “Add a tutorial on inviting a teammate, then check it on this branch.” Your agent:

1. reads the Ondaclip docs and the tutorial format;
2. finds the screens and buttons for that flow in your code;
3. writes the tutorial file: each step, its action and its narration;
4. checks it with the Ondaclip MCP server and fills in anything missing;
5. opens a pull request.

Then a person reviews the videos in the studio before anything is published. The agent does the prep work. Your team makes the call.

Why an agent beats a hand-written script

A hand-written tutorial is only as good as its author’s memory. An agent starts from the code: it finds the right identifier for each button, offers to add one when it’s missing, and writes the narration in all your languages. When a screen changes, it rereads the code and updates the tutorial in the same pull request as the change.

Set up the Ondaclip MCP server once

Three steps, once per developer:

• create a personal token in the studio;
• add the Ondaclip MCP server to your agent with a single command;
• paste a short snippet into your repo’s AGENTS.md, so every agent knows where your tutorials live.

It’s all covered in your AI agent. If you’re starting from scratch, the quickstart sets up the .ondaclip/ folder in your repo first.

Stay in control of what gets published

The agent never gets access to your demo accounts, and it can’t publish. Every tutorial goes through a pull request, then a video review. To keep those tutorials accurate over time, read why your video tutorials age faster than your product.