Atelier201 websiteJoin the waitlist
PlatformGuest sites

Guest sites

An agent prepares a site before its owner has an account, and the owner claims it.

A guest site lets an AI agent build a site for someone who has no Atelier201 account yet. The agent prepares a draft and a demo link. The person looks at the demo, signs up and claims the site. Nothing is published, no domain is connected and no credits are spent along the way.

The flow

  1. The agent prepares the site with Prepare a site before sign-in. The reply holds the site_id, a write_token that works for 2 hours and a claim_url that works for 7 days.
  2. The agent sends its project as a tar archive (import), or keeps the Atelier201 starter, then builds it. A guest site gets 3 builds.
  3. The agent polls the site’s state until demo_url appears, and shows it to the person. The demo works for 7 days.
  4. The agent gives the person the claim_url in a private message.
  5. The person opens it, signs up or signs in, and the site becomes theirs.
  6. If the agent asked for it, the person chooses whether to connect that agent. Declining keeps the site.

From the CLI, the same flow is one command:

Terminal
atelier201 site push --guest --name "Maison Dupont"

It prints the demo link and a private link for the owner, and keeps the write token in .atelier201/site.json without ever printing it.

What a guest cannot do

A guest site can be previewed, not published. The demo runs in a sandbox, is not indexed and has no forms. The write token opens that one site and nothing else, and it stops working once the site is claimed. A guest project is limited to 10 MB and 500 files.

An unclaimed site and its demo expire. Prepare the site again if the link ran out.

Keep the secrets private

The write_token and the claim_url are capabilities. Keep them out of shared chats, logs and commits. Whoever holds the claim_url can claim the site.

After the claim

The site is a normal site of the person’s workspace. To publish it, the person (or an agent they connected, with sites:publish) reviews the build and publishes it. The claim and the connection are not that approval. See How a site goes live.

Use these docs with your AI tools

An AI agent can read this documentation directly. You do not need an account or an API key. Everything here is public and read-only.

Query these docs via MCP

Recommended

Add this server to Claude, Claude Code, Cursor, Mistral, or any tool that supports MCP. Your agent can then search Atelier201 Docs documentation and read it in full, instead of answering from memory.

https://atelier201.com/docs/mcp
  • searchFind the passages that answer a question.
  • fetchRead one page in full, as Markdown.
  • list_pagesSee every page in this documentation.

Query these docs over HTTP

The same tools also work as plain web requests. Use this for scripts, or for any tool that does not support MCP. There is one endpoint per tool. Arguments go in the query string, and the answer comes back as JSON.

https://atelier201.com/docs/api/docs/search?query=custom+domain

Read the OpenAPI description. It is built from the same definitions as the tools, so it always matches what the endpoints do.

Read these docs as Markdown

Add .md to any page URL to get its Markdown source. You can also send the headerAccept: text/markdown to the page URL itself.

To read the whole documentation in one file, open llms-full.txt. For a short index of every page, open llms.txt.