Transactional

Developers who need to generate PDFs like invoices or receipts today often run their own rendering stack, typically Chromium via Puppeteer in Lambda. This means dealing with cold starts, out-of-memory crashes, missing system fonts, and maintaining a fragile render pipeline themselves. Designing the actual document layout also requires touching a layout engine or hand-coding HTML/CSS for each template. Updating a template means redeploying code rather than just editing a design.

Transactional lets a user design a PDF template once in HTML and Tailwind, either by writing it themselves or having an AI agent generate it, using Handlebars placeholders, conditions, and each-loops to mark where data should land. The backend then calls one API endpoint with a document id and variables to generate a personalized, pixel-stable PDF, which is compiled, rendered, and hosted on their CDN with a signed URL returned. The whole API is also exposed through MCP, so an MCP-aware LLM (Claude, ChatGPT, Cursor, Gemini) can create documents, edit HTML, render test PDFs against sample data, and write the integration code directly into a codebase. Every call is logged in a dashboard with status, latency, and payload, and the system is built to scale from one PDF to thousands without the user managing rendering infrastructure.