Docs have a second audience now.
And it's the larger one.
Somewhere in 2026 the share of documentation traffic coming from AI agents passed the share coming from browsers. People stopped searching the docs and started asking Claude, Cursor and ChatGPT — and those tools went and read the docs for them.
Every documentation platform responded the same way: make the docs easier for agents to read. Markdown at every URL, llms.txt, an MCP server. That is table stakes now, and Orbizz does all of it.
What nobody did
Nobody asked what the agent could give back. An agent that searches your docs and finds nothing has just told you exactly which question your docs don’t answer. An agent that reads a rate limit of 1,000 and gets a 429 at request 612 has just found an error no human reader would report. That signal was being thrown away.
Orbizz keeps it. Agents report the gaps. Agents propose the fix, with evidence. A person approves, and the page is published and versioned. The docs learn from the readers that read them most.
Who it’s for
Small teams first. The ones without a support team, where the founder answers the Discord and the docs go stale because nobody has time. For them, agents answering from the docs and reporting what’s missing isn’t a feature — it’s the support team.
How it’s built
Self-hosted, one database, three audiences. This site is served by it: every page under /docs, /blog and /changelog is an Orbizz document, and the playground queries them through the same functions the MCP server exposes. We would not ask you to trust a product we don’t run ourselves.
Orbizz is in beta and free. If you want to be one of the first teams on it, we’d like to hear from you.
Start writing for both readers.
Free while in beta. Self-hosted, your content, your database.