Blog

Markdown in, designed PDFs out.

Playbooks for turning what your AI agent writes — proposals, reports, runbooks — into consistent, print-ready documents.

5 min read

Extract tables from a PDF to JSON or CSV — without letting an AI invent cells

One API call returns every recoverable table in a PDF as rows and CSV, tagged with page provenance. Deterministic first — a table that can't be recovered honestly comes back as a warning, never a half-guessed grid.

Read post
5 min read

How to fill PDF forms automatically with AI (from JSON, a spreadsheet, or Claude)

Government filings, insurance claims, onboarding packets — the answers are already structured data. Two API calls put them on the official form: inspect the fields for free, then fill them. No Acrobat, no browser bot, no silent drops.

Read post
5 min read

Documents in, structured data out, styled documents back

Inkrun now extracts structured, grounded data from PDFs — invoices, receipts, contracts — through the same MCP server and API that render your agent's PDFs. One key, one credit meter, both directions.

Read post
5 min read

Why your AI agent shouldn't write CSS

LLMs are excellent at content and unreliable at design. Giving an agent raw HTML/CSS control creates drift and a security hole. The fix is an old idea: separate content from form.

Read post
6 min read

Markdown to PDF: every way to do it in 2026 (an honest round-up)

Pandoc, VS Code extensions, md-to-pdf, Puppeteer, HTML-to-PDF APIs, and hosted services — what each is actually good at, and how to pick for your case.

Read post
4 min read

How to get a real PDF out of Claude (5 minutes, no code)

Connect the Inkrun MCP server to Claude Desktop or claude.ai and turn any conversation into a designed, print-ready PDF — no copy-pasting into Google Docs.

Read post