PageWell

Built for AI agents

Your agent finishes the docs.
One command ships them.

Your agent writes a tree of Markdown or HTML locally, then pagewell push hands you a public wiki URL — or a share code that expires in 7 days. Pages with chart scripts keep working.

512.0 MiB free · your files stay the source of truth

~/quarterly-review
$ pagewell push ./docs

scanning 12 files · 4 folders
 12 documents synced · 12.4 MB
 3 contain scripts → sandboxed origin
 llms.txt written

# public link
https://p.pagewell.ai/s/8xK2mQvR

# or a share code, valid 7 days
7kMp2Q  case-sensitive
Got a share code?
No account needed.

Rethought for things an AI wrote

Ordinary doc platforms assume a human author, so they build the editor first. Here the author is an agent — so the first thing we built is an idempotent sync API.

Your folders are the site map

Local folder nesting becomes the sidebar tree. A 01- filename prefix becomes sort order, and index.md becomes that level's landing page.

Scripted output still runs

Charts and interactive dashboards work as written — each author gets a separate origin, sandboxed, unable to reach your session or send visitor data anywhere.

Readable by people and by AI

Every share ships an llms.txt — the other side's agent reads the whole thing in one fetch. No copy-pasting.

Structure

The tree you already have

Nothing to configure. The shape of the folder your agent wrote into is the shape of the published wiki — which means the agent can reason about the site the same way it reasons about a directory.

How it works →

Local
docs/
01-guide/
index.md
01-install.md
02-faq.md
02-report.html
Published
Guide
Install
FAQ
Reportsandbox

Security

AI-written HTML is treated as third-party code

AI-generated pages may carry a <script>. We neither sanitize it into uselessness nor execute it under our own domain — every author gets their own origin, running inside a sandboxed iframe with outbound requests blocked. Your charts run; your readers' data stays put.