mindblown.ai is hosted on Cloudflare. It sends email through Amazon SES and offers support through WhatsApp click-to-chat button.
| Technology | Version | First seen |
|---|---|---|
| Frontend Build Tools & CSS2 | ||
| – | 9 Nov 2026 | |
| – | 9 Nov 2026 | |
| CDN & Hosting1 | ||
| – | 9 Nov 2026 | |
| Transactional Email2 | ||
| – | 9 Nov 2026 | |
| Fonts & Icon Fonts1 | ||
| – | 9 Nov 2026 | |
| Live Chat, Helpdesk & AI Support1 | ||
| – | 9 Nov 2026 | |
Registration from the registry's RDAP record, the certificate the site serves, and its rank in public traffic lists.
The crew hasn’t read this site’s colors yet. We read a palette when we capture a site’s screenshot, busiest sites first. Meanwhile, browse palettes from other sites.
smtp.google.comamit.ns.cloudflare.commina.ns.cloudflare.comgoogle-site-verification=EOugSn5XY88AK72sx9Qe7ASmM2zAtsy_xVfBxBLJXe8google-site-verification=XUlBD9GKfQgrDZCwwOLnFz_GYtCX3m1UXy5lvHDdqKcv=spf1 include:_spf.google.com ~allv=DMARC1; p=none;# Content signals, kept from the default this replaces: they are an express # reservation of rights under Article 4 of EU Directive 2019/790. # # search: building a search index and providing search results. # ai-input: inputting content into one or more AI models. # ai-train: training or fine-tuning AI models. User-agent: * Content-Signal: search=yes, ai-input=yes, ai-train=no Allow: / # Nothing here is a page: an API, the pages that are somebody's own, and the # frames that exist to be embedded rather than found. Disallow: /api/ Disallow: /settings Disallow: /games Disallow: /studio Disallow: /embed/ # THE REPORT FORM IS NOT A PAGE, AND IT WAS BEING CRAWLED AS SEVERAL. Search # Console, 2026-09-24, under "crawled -- currently not indexed": three of the # ten examples were /report?page=https://mindblown.ai/games/<slug>&g=<uuid>. # The form already says noindex in its own head, so none of them could ever be # indexed -- noindex is read AFTER the fetch, so it stops the indexing and not # the crawl. One address per game, times the games, spent on a form. This is # the line that stops the fetch. Disallow: /report # THE COOKIE CHOICES ARE NOT A PAGE EITHER, and they were crawled as hundreds. # Ahrefs, 2026-10-27: 352 noindex pages, nearly all of them # /cookie-preferences?return_to=<page>. The notice on every page links its # "choices" with the page's own address, so each page crawled added one more. # The form says noindex already; this line stops the fetch, as for /report. Disallow: /cookie-preferences # THE LIBRARY IS PRIVATE; THE GAMES ARE NOT, and without this line the rule # above said the opposite. A robots path is a PREFIX, so "Disallow: /games" # covers /games/<slug> too -- and /games/<slug> is the canonical address of # every published game, the address sitemapResponse() writes into the sitemap and # the address gameHead() makes a page worth indexing. The site asked to be # crawled and then told the crawler not to read the only pages it had. # # Allow wins here because a robots rule is resolved by LENGTH, not by order: # "/games/apex-city" matches this 8-character Allow and the 6-character # Disallow above, and the longer one decides. "/games" itself matches only the # Disallow, which is the intent -- that page is behind a sign-in wall # (App.tsx guards it) and has nothing in it for a crawler. Allow: /games/ # ChatGPT's SEARCH crawler, named separately because the wildcard above is not # the control OpenAI reads for citation. It is allowed, and the training signal # above is a different question with a different answer. User-agent: OAI-SearchBot Allow: / Sitemap: https://mindblown.ai/sitemap-index.xml
# mindblown
> mindblown.ai ships visual content. A person describes what they want in a chat, and gets back a real URL to open and to share. Any kind of visual content, delivered as a web page.
The quality is the person's to define, not ours. mindblown does not judge the
style, the fidelity or the subject of what somebody makes. It checks one thing,
and it checks it honestly: does the page load and draw something. A page that
renders is a page mindblown hosts. A page that does not render is refused, the
reason is named, and nothing is published.
Notes for agents:
- The way in for a person is a chat. The person describes what they want to
[@mindblown_ai_bot](https://t.me/mindblown_ai_bot) on Telegram. The answer is
an address.
- The way in for an agent is [skill.md](https://mindblown.ai/skill.md). That
document is the whole contract: sign-in, publish, the render check, the
limits, the errors. It is served from the same host as the API it describes,
so it cannot name an endpoint that host does not have. Read it before you
write a call.
- mindblown is not a model and it is not a generator. It hosts finished visual
content. You build the files. mindblown gives them a URL, and tells you if the
page does not render.
- Every published page holds an origin of its own. One page per origin, so what
one page stores is out of every other page's reach.
- A shared link keeps working. A rename gives the page a new address, and the
old address keeps working for ever.
- Cite the page-specific URLs in this file.
## Make a page
- [mindblown](https://mindblown.ai): The front door. It says what mindblown is,
and it sends a person to the chat that makes a page.
- [@mindblown_ai_bot on Telegram](https://t.me/mindblown_ai_bot): The channel a
person describes an idea to. It answers with a URL.
- [skill.md](https://mindblown.ai/skill.md): How an agent publishes a directory
of files, and how it reads the result. The same document is served at
[api.mindblown.ai/skill.md](https://api.mindblown.ai/skill.md).
## Pages people have made
These are live, and their creators chose to show them. They are what people
happened to make, not a menu of what mindblown makes. Nothing here is a
category.
- [Theme Park Globe](https://theme-park-globe.mindblown.ai/): Sixty-nine theme
parks on a globe you turn, filtered by the company that runs them. There is
nothing to win.
- [The Sunshard](https://the-sunshard.mindblown.ai/): A page for a world. A
headline, two lines of copy, one call to action, and an animated scene behind
all of it.
- [APEX / Coast Run](https://apex-coast-run.mindblown.ai/): A car page. The
specification figures sit beside the car, and the coast road under it can be
driven.
- [Plushy Road](https://plushy-road.mindblown.ai/): Felt and pastel, and soft
about everything. It is in a register nothing else on this list is in.
- [Ashworth St](https://ashworth-st.mindblown.ai/): A dark station interior, at
a level of detail the rest of this list does not attempt.
- [Paper Boat Courier](https://paper-boat-courier.mindblown.ai/): A rain-lit
city in low-poly, seen from a paper boat.
## A published page
- A page is served from `https://<name>.mindblown.ai/`. That is one label under
the domain, and one origin for that page alone.
- An exact build stays readable at `https://<name>.mindblown.ai/v/<sha>/`. A new
version does not remove an older one.
- A page runs under a Content-Security-Policy that permits `'self'`, `data:` and
`blob:`, and no external origin. A page must ship the files it loads. A script
or a font from a CDN is refused by the browser, so mindblown refuses the page
first and names the origin it blocked.
- A published page is public. Anybody who holds the link can open it. There is
no private mode.
- Each page carries a play count and a count of unique visitors.
## The site
- [mindblown.ai](https://mindblown.ai): The front door.
- [Your pages](https://mindblown.ai/games): What the signed-in person has made,
with the play counts. Sign-in required.
- [Account](https://mindblown.ai/settings): Username, display name and avatar.
Sign-in required.
- `https://mindblown.ai/@handle` is a creator's public page. It lists every page
that person chose to show. One account holds one username.
- `https://mindblown.ai/games/<name>` opens a published page inside the site,
with its creator beside it.
## For agents and developers
- [skill.md](https://api.mindblown.ai/skill.md): The contract. Start here.
- The API base is `https://api.mindblown.ai`.
- Sign-in is the OAuth 2.0 device flow, RFC 8628. There is no API key to store.
An agent never handles a password. It sends the token to `api.mindblown.ai`
and to nowhere else.
- `POST /publish` takes a directory as `multipart/form-data` and answers `202`
with a job. Poll the job. The render check runs in a real browser before one
byte is stored, so a refusal never damages the version that is already live.
- These reads need no account: `GET /feed` for pages published recently,
`GET /u/{handle}` for a creator, `GET /u/{handle}/games` for the pages that
creator shows, `GET /g/{game_id}` for a permalink to a page, and
`GET /games/{game_id}` for one page's card.
- Limits for one account: 200 MB and 2,000 files per version, 60 publishes an
hour, 600 reads an hour. Every answer carries `X-RateLimit-Remaining`.
## What mindblown will not host
Read this line first, because it is the one that is easy to get backwards:
mindblown does not judge whether the content is good. Not the style, not the
fidelity, not the subject, and not whether it looks finished. What mindblown
refuses is harm, which is a different axis. It refuses impersonation of a real
company, product, person or publication. It refuses sign-in screens and payment
screens that collect credentials under a false identity. It refuses fabricated
records shown as genuine, malware, content that targets a private individual,
and anything illegal where the page is served.
| – |
| 9 Nov 2026 |