Show HN: Papermill Press – An AI-friendly markup language for PDF generation
25 comments
Hi Folks - I'm David, CEO of Papermill. We're super excited to get feedback on what we're building. This is the first time we've posted to a forum. Designing a document language for AI is a major undertaking and we'd love to hear from folks with experience and interest in document languages and generation.
As a solicitor myself managing a high street law firm that is already comfortable drafting, reviewing, and producing client-ready PDFs from Word, where is the return on investment in adopting Press? What can Press do that cannot be achieved with Word and existing document automation tools?”
Papermill is more the "infrastructure" layer that sits below legal AI tools - we don't work directly with law firms, rather more likely to be used by a startup building a legal tool. Companies like Wordsmith, Legatics, Legora, Harvey can use Papermill to output high-quality PDFs.
That said, an export from Word to Papermill templates might be a nice addition...
That said, an export from Word to Papermill templates might be a nice addition...
It's a great tool. I was trying to create a document with Claude, and somehow half of the document was completely omitted without me noticing. Hence, I was looking for a tool dedicated to document creation/formatting.
I found a particular issue in the market around formatting mathematical formulas. Do you guys already have a solution for that?
I found a particular issue in the market around formatting mathematical formulas. Do you guys already have a solution for that?
Hi, Tom here. I'm happy to answer any questions r.e., generating PDFs, better ways of making automated document workflows more robust, or anything related to Papermill and Press!
I'm excited to see more about how Papermill and Press can interact with agents in agentic workflows! This seems like a great problem solved in AI-first startups.
Press is designed from the ground-up for agentic workflows. We spoke to hundreds of companies generating PDFs before we designed the language, and had been building AI design systems ourselves for a few years.
You can connect an agent to our MCP server, see: https://docs.papermill.io/mcp/#using-the-papermill-mcp-serve...
It's quite amazing to see what Claude can do when it has a tool like Papermill at its disposal - worth connecting and having a play.
In practice, you can either directly connect an LLM to the MCP server, or send via the API after (say) cleaning up LLM output, combining it with RAG and other data etc.
You can connect an agent to our MCP server, see: https://docs.papermill.io/mcp/#using-the-papermill-mcp-serve...
It's quite amazing to see what Claude can do when it has a tool like Papermill at its disposal - worth connecting and having a play.
In practice, you can either directly connect an LLM to the MCP server, or send via the API after (say) cleaning up LLM output, combining it with RAG and other data etc.
Yep, I've been experimenting today!
Great! I’ll be posting videos on social media of MCP use over the next few weeks. Be sure to follow @papermillapp and keep an eye on our blog.
Thanks Ste, we're happy to answer any questions you've got!
This looks cool - I’ve been building pdf rendering for agents this year and styling is such a pain. I’ll check it out
Thanks Jonny. What were you rendering?
I'd love to see something like this in chatgpt or claude. The PDFs you can generate there are always so boring.
Thanks Tom, good to see you on HN! You can connect Papermill to your favourite agent via MCP - we prefer Claude Code.
https://docs.papermill.io/mcp/#using-the-papermill-mcp-serve...
You can ask Claude to generate templates, full documents, ask it about the Press language, save templates, ask it about templates and more. There's a lot more to it - design guidelines for print vs web, recipes in Press etc.
I've tried it with Fable already and it's a noticeable improvement - we support visual feedback through MCP and I think that's helping Fable a lot.
https://docs.papermill.io/mcp/#using-the-papermill-mcp-serve...
You can ask Claude to generate templates, full documents, ask it about the Press language, save templates, ask it about templates and more. There's a lot more to it - design guidelines for print vs web, recipes in Press etc.
I've tried it with Fable already and it's a noticeable improvement - we support visual feedback through MCP and I think that's helping Fable a lot.
Hi Tom,
One of the things we've learned is that models often need more guidance when generating documents. We give real-time feedback to models as they're constructing the document, such as "you have content here that you've done nothing with and doesn't appear in the report as it likely overflowed off the page, you should think of a way to handle this" and "this image is sized incorrectly and just will not fit in the area you've specified".
Giving models the tools to create media for print allows both the user and the model to experiment with more wild designs, knowing that things will not randomly break!
One of the things we've learned is that models often need more guidance when generating documents. We give real-time feedback to models as they're constructing the document, such as "you have content here that you've done nothing with and doesn't appear in the report as it likely overflowed off the page, you should think of a way to handle this" and "this image is sized incorrectly and just will not fit in the area you've specified".
Giving models the tools to create media for print allows both the user and the model to experiment with more wild designs, knowing that things will not randomly break!
Looks interesting. Are there any limits on the file size or length of pages that can be produced?
Hi off_by_two
We don't apply hard limits on PDFs. We've handled PDFs of over 1GB and close to 1000 pages.
We also have a large document service that handles extremely large PDFs in the background.
We don't apply hard limits on PDFs. We've handled PDFs of over 1GB and close to 1000 pages.
We also have a large document service that handles extremely large PDFs in the background.
Worth adding that we also support the integration of many PDFs into a large document, including complex processing, but this is not part of the publicly accessible Press language (yet).
looks great dave! loved the product. Hope it all goes well.
V cool. Glad to see another Manchester based start up!
Byteful seems cool! WEb proxy stuff is important for me in my near future so I may just check you out. Thanks for commenting!
Thanks Tim! Great to see Byteful doing so well too!
The fundamental problem is that HTML was designed for screens, not print.
We built Press, a markup-based document language where pages, content flows, and assets are first-class concepts. Content can flow across frames, columns, and pages without manual pagination. Pages are created dynamically based on the available content.
Press templates separate layout from content. You can send markdown, Press markup, or a mixture of both to the API. Data can be sent in JSON, CSV, XML.
Because Press is XML-based it can easily be generated by agents - some of our users are generating complete documents in a single shot, although the language is designed for repeatable automation.
You can also use our MCP server, which enables models to design templates.
A simple API call sends a markdown payload, which is injected into a <flows><body>…</body></flows> element in Press:
PDF output: https://mill.pm/hn
Pages and frames in Press can declare dependencies on a content flow like the <body> above. By default, if there’s no content in the flow then the frame or page won’t be generated. You can run flows between frames and pages, and combine multiple flows on a page - for example, a sidebar can run across pages until no content is left, then make room for body content. This makes it possible to implement complex layouts.
You can mix markdown and Press:
The typesetter adapts to dynamic content (e.g. LLM output). For example, tables and columns can be automatically sized and Papermill will even auto-rotate a table and its page to fit if needed.
Templates support components, repeating over data, document logic, and conditional styling. We mostly use an inline-styling approach, and provide the concept of a style “alias”, which is a bag of styling properties you can reuse and compose.
Here’s an example template written in Press, our document language. It uses the first page layout until the sidebar flow is exhausted, then switches to the second:
Papermill is a paid API with a free tier. Press is the document language.
Try it for free: https://app.papermill.io/signup (no credit card needed)
Docs including MCP setup: https://docs.papermill.io
Data-only sandbox: https://app.papermill.io/demo.html (no email needed)
We're a small team based in Manchester, UK. Tom (CTO) and I are happy to answer questions about the language design, the rendering engine, or anything else!