At PaperCast, we believe in dogfooding our own tools. When we wanted to test the limits of our framework-agnostic Pagination Engine and strict JSON schema, we decided to let an AI agent build a complex 12-page document entirely on its own.
By exposing our builder through the Model Context Protocol (MCP), we gave the AI agent safe, surgical tools to manipulate the AST without ever touching the raw JSON file directly.
The Process
Instead of writing a massive JSON payload by hand, we provided the AI agent with a series of conversational prompts. The strict typing of our schema meant the AI couldn't hallucinate invalid properties.
Here are some examples of the prompts we used:
Prompt 1: "Add a cover page with a large title 'Annual Report 2026', a subtitle, and a full-width hero image."
Prompt 2: "Create a complex Table of Contents with dotted leaders connecting the chapter titles to the page numbers. Use a
RowwithjustifyContent: 'space-between'."
Prompt 10: "Add a dynamic Table footer that calculates the sum of the 'amount' column, and merge the cells to display 'Grand Total'."
Prompt 12: "Create a new page detailing a fun fact about this capability: these pages are built entirely with the MCP tool and playground."
Because the AI agent had access to tools like insert_layout_element, add_text_widget, and search_icons, it constructed the document iteratively, measuring and paginating in real-time.
The Result
The result is a beautifully formatted, auto-paginated 12-page document. Our Pagination Engine handled all the complex page breaks, header injections, and footer calculations automatically.
Below is a lightweight preview of the document, powered by our WidgetCard component (which uses shiki for syntax highlighting, rather than a heavy Monaco editor).
View the Document Preview & Schema
Why Strict Schemas Matter
If we had allowed the AI to generate raw React code or raw HTML, it would have been impossible to guarantee that the document could be paginated accurately for PDF export.
By constraining the AI to our strict DocumentSchema (and enforcing updates via our AST manipulation tools), every element remained perfectly measurable by the OffscreenMeasurer.
Try the Real-Time Embedding Example
We recently launched a powerful new feature that allows third parties to embed the PaperCast playground via an iframe on their own sites, enabling seamless cross-origin schema injection via the window.postMessage API.
Want to see it in action? Check out our Real-Time Embedding Example where you can modify a JSON schema in real-time and watch it instantly update the embedded playground!