A folder of documents is already a website.
Notes you wrote, or a report your AI did — drop the folder, or just the one file. No index.html, no theme, no front matter, no generator: the files are served through a reader that builds its sidebar from the folder itself.
One stepnothing to install
Drag the folder onto the home page and the link comes back when the upload finishes. From a shell or an agent instead: the CLI and the MCP server take the same folder.
Readable on arrival, not after configuring it.
A folder with no index.html and at least one document — markdown, PDF, Word, Excel or EPUB — is served this way. Both are the same publish; what decides is what is in the folder. How it works →
- sidebar
The folder is the navigation
The tree comes from the real directory, so nothing describes the structure twice. No sidebar file to write and no order to maintain by hand.
- rendered
Code, maths and diagrams
Syntax highlighting, KaTeX maths, Mermaid diagrams and GitHub-style alerts render as themselves — which is most of what a document written by a tool contains.
- .pdf .docx .xlsx .epub
Files read in place
A PDF, a Word document, a spreadsheet or an ebook is rendered rather than downloaded, so a link to one opens on a page instead of in somebody's downloads.
- one file
One file is enough
A single report or a single PDF is a whole publish. It opens straight into the reader, with no folder around it and no sidebar to scroll past.
- curl
One link, two readers
A person opening the link sees the reader. An agent or
curlfetching the same address gets the file itself, so the link you share is one a model can read too. - theme
Light, dark, raw
The reader follows the visitor's theme, and every document has buttons for the raw file and a download — and, for anything that is text, a copy.
No generator, and nothing to learn first.
The folder goes up as it is. What is below is not a shorter setup than a static-site generator — it is the absence of one.
- ✓No
index.htmlneeded. A folder of documents publishes with no entry point at all. Put one in and it wins — the folder is then an already-built site, which is the other thing this hosts. - ✓No front matter, no config, no sidebar file. Nothing in the folder has to declare a title, an order or a layout. Files a tool wrote for a human to read publish exactly as they came out of it.
- ✓Relative links and images resolve on their own. A document keeps its real path, so an image beside a note and a link to the file next to it both work without anything rewriting them.
- ✕The default subdomain is not indexed. Every response on it carries
X-Robots-Tag: noindex, deliberately: a folder of working notes is shared with people, not published to search engines. A custom domain does not carry that header — see the plans. - ✕It is a reader, not a wiki. Nothing is edited in the browser and there is no search across documents yet. Publishing again replaces what is served, and the previous version stays available to roll back to.
What a chat writes is already a document.
Claude and ChatGPT answer in markdown — headings, tables, code blocks, often a Mermaid diagram or a formula. That is exactly what this reader renders, so a report copied out of a conversation reads the way it did in the chat, on a link anyone can open.
Paste it into the box on the publish page, or let the assistant publish it itself through the MCP server — the same document either way, no file saved first.
And when the people reading it have something to say, they can say it on the document: on plans that include comments, a reader selects a sentence in a Markdown, code or text file and comments on it. The assistant reads those threads, revises the file and answers on the page. How comments work →
Enough for a folder of documents, free.
The free tier's figures, read from the same table the pricing page reads. Markdown is text, so the file count is the limit a folder of notes meets first — not the bytes.
2,000 files per publish
Documents, images beside them, and PDFs, all counted together after the junk is dropped.
10 MB per file
Which a document never reaches and a scanned PDF sometimes does. It is the figure to check before dropping a folder of them.
50 MB per site
Everything one site holds, across every version of it that is still kept.
7 days without an account
Publishing needs no account at all, and those sites expire. Signing in makes one permanent, and the same link claims a site published without one.
Drop the folder, send the link.
Notes, a report, a specification, a folder of PDFs. Nothing to build and no account needed to start.