Documentation that your users actually find
Stories, deep dives, and field notes from the team building the open-source, AI-ready documentation platform.
Latest stories
14 postsDocumentation drift: how to detect it and fix it automatically
A page can be edited three times this month and still be wrong, because freshness and accuracy are different measurements. Most docs stacks can only take the first one. Here is what it takes to measure the second, and how much of the fix can safely be automated.
llms.txt: what the spec actually says
Almost every llms.txt I open is doing too much and missing the one thing the spec actually recommends. The file itself takes four lines. The part people skip is the part an AI agent actually fetches.
How to password protect a documentation site
Putting a login screen in front of your docs is the easy part. The part teams miss is that a documentation site has four separate doors — pages, search and chat APIs, crawlers, and the machine-readable files agents read — and a password on the front door leaves three of them open.
How to generate a documentation site from a GitHub repo
Most codebases have no real documentation because writing it from a blank page is the part nobody gets to. Here's how to go the other direction: point an AI at your GitHub repo, have it read the actual code, and get a complete, editable docs site as a first draft you refine — instead of a blank docs/ folder you never fill.
How to write API documentation developers actually use
Most API documentation fails the same way: it describes endpoints instead of helping someone ship. A practical guide to structuring API docs — the five page types every API needs, what belongs on a reference page, and how to write for the AI agents that read your docs as often as humans do.
How to create a documentation site from Markdown files
A practical guide to turning a folder of Markdown files into a real documentation site: what a docs site actually needs in 2026, the decisions that matter before you pick a generator, and a path from zero to a live site in one command.
How to keep your documentation up to date
Stale docs are the most common documentation problem and the easiest to ignore. Six patterns that keep documentation fresh without doubling the team that writes it.
Documentation without lock-in: how to keep your content portable
If your docs live inside a proprietary editor, you don't own them — you rent access to them. Here's what real portability looks like: Markdown in a Git repo you control, a build you can run yourself, and the right to walk away without a migration project.
How to connect your docs to Cursor, Claude Code, and other AI coding tools
Your users build with AI coding agents now. If those agents can't read your current docs, they guess — and ship the guess into production. Here's how an MCP server fixes that, and why it should already be part of your docs site.
How to track AI bot traffic on your documentation site
GPTBot, ClaudeBot, PerplexityBot, and a long tail of agents are reading your docs every day. A practical guide to spotting them in your logs, measuring what they actually do, and fixing the gaps before they ship into a developer's editor.
Documentation is the new SDK
AI agents have made documentation a runtime concern. The contract between your product and the developer ecosystem used to live in your SDK. Increasingly, it lives in your docs.
How to write documentation that AI agents can actually use
AI assistants now read your docs more often than humans do. A practical guide to the structure, formatting, and metadata patterns that make documentation parseable, retrievable, and citable by LLMs.
How to add dark mode images to your documentation site
Doccupine renders separate images for light and dark mode using one line of MDX. No imports, no custom components, no flash of the wrong asset on theme switch.
Welcome to Doccupine
Doccupine ships your documentation at the speed of your code — versioned in Git, deployed automatically, live on your own domain in minutes.