# Super Artifacts

> A hosting and data plane for artifacts an AI agent builds. The agent deploys a small web app to a live URL, people use it, and what they enter flows back to the agent over MCP. The platform runs no model; your agent writes every byte.

Human site: https://superartifacts.app
Documentation: https://superartifacts.app/docs.md
Skill file: https://superartifacts.app/skill/SKILL.md
Every page as markdown, one index: https://superartifacts.app/llms.txt

## What it is

Your agent builds it in a folder and ships it to a live URL in seconds. Real people open it, use it, and every answer flows back to the agent that built it.

An artifact is deployed at a URL with a database behind it, so real people can open it on their phones, use it, and have what they typed flow back to the agent that built it. It is the same idea as an artifact rendered inside a chat window, plus the half those tools stop at: the data loop.

The platform generates nothing. Every byte comes out of your agent's model, on your tokens, in your environment. There is no endpoint that takes a prompt and returns a page.

## What makes it different

- **Real data, retained the right way.** Every artifact is backed by structured data, not a static page. What an artifact may collect, and who may read it, is a policy the deploy declares, enforced per row. The agent reads rows back with `super_data` and `super_data_query`.
- **Works with whichever agent you are using.** An artifact is not tied to the agent that built it. Any MCP-speaking agent can read it and pick up the work.
- **Templates save tokens.** Starting from a template instead of a blank page means less back-and-forth generation. Public templates are listed at `https://superartifacts.app/t/{slug}`.
- **Playlists keep artifacts organized.** Group related artifacts into a playlist with `super_playlist`.

Artifacts are private by default. Sharing is an explicit grant: named people, or public to anyone with the link. Every deploy writes an immutable version, and rolling back is the same operation as rolling forward.

## Connect an agent (MCP over OAuth)

- Endpoint: `https://api.superart.page/mcp` (HTTP POST)
- OAuth authorization server metadata: https://superartifacts.app/.well-known/oauth-authorization-server
- OAuth protected resource metadata: https://api.superart.page/.well-known/oauth-protected-resource
- Scope: `artifacts:write`. PKCE with S256. Dynamic client registration at `https://superartifacts.app/oauth/register`; the device flow (RFC 8628) starts at `https://superartifacts.app/oauth/device_authorization`.
- Claude Code: `claude mcp add --transport http super-artifacts https://api.superart.page/mcp`
- Without OAuth, send a key in an `Authorization: Bearer` header.

Harnesses with a documented connect path: Claude Code, Claude Desktop & claude.ai, ChatGPT, Codex, Hermes, OpenClaw, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Zed, Devin, Grok Build, Antigravity, OpenCode, Goose, Cline, Amp, Kiro, Junie, Warp, Qwen Code, Kilo Code, Manus. [Connect an agent](https://superartifacts.app/docs/connect-an-agent.md) has the details, and [Agents](https://superartifacts.app/features/agents.md) lists the commands.

## Tools (16)

- `super_whoami`
- `super_publish`
- `super_artifacts`
- `super_artifact_set`
- `super_data`
- `super_data_query`
- `super_data_schema`
- `super_entries`
- `super_templates`
- `super_template_save`
- `super_template_action`
- `super_release_template`
- `super_playlist`
- `super_share`
- `super_comments`
- `super_help`

Every answer is one JSON envelope: `{ "ok": true, ... }` on success, or `{ "ok": false, "code", "message" }` on failure.

## First calls

1. `super_whoami` confirms the account before anything is published.
2. `super_help` returns short usage notes by topic, fetched on demand.
3. `super_publish` deploys files at a stable URL. Without a `policy` the artifact stores no visitor input; add one if it should collect data.

## Pricing

Free to start. Solo is $19 a month or $190 a year, with a 14-day trial. Business is coming. Details: [Pricing](https://superartifacts.app/pricing.md).

## Questions

### What is a super artifact?

Something your agent builds and deploys to a URL people open — a tracker, a survey, a playlist. Not a transcript you have to explain, and not a code block somebody has to run: a page.

### How is this different from Claude or Codex artifacts?

Those render something in a window you look at. A super artifact is deployed at a URL with a database behind it, so real people can open it on their phones, use it, and have what they typed flow back to the agent that built it. It is the same idea, plus the half those tools stop at: the data loop.

### How does an agent get paired?

One MCP URL. Point the harness at it — `claude mcp add --transport http super-artifacts https://api.superart.page/mcp` — the harness offers a Connect button, the consent screen names the account, and the key stays in the harness where you can revoke it.

### Where does the data go, and who can read it?

Into a database that comes with the artifact. What the artifact may collect and who may read it is a policy the deploy declares, enforced per row: an artifact that says a visitor reads only their own entries cannot hand them anybody else’s, and an artifact marked private serves nothing to somebody who has not been named.

Erasing a visitor erases their rows on both the serving plane and in the database, and deleting an artifact closes the whole surface behind it.

### Does the platform generate any of the artifact?

No. Every byte comes out of your agent’s model, on your tokens, in your environment. There is no endpoint here that takes a prompt and returns a page, and there is not going to be one — an agent that expects the platform to generate the artifact will post a prompt and get back an empty page.

### What is not built yet?

There is no per-artifact analytics report yet — the view and error pipelines are live and the gallery shows counts, but the page that breaks them down is not built. There is no `super` command-line tool, only the HTTP endpoint it would wrap. Custom domains are not built either, which is why they are not listed as a Solo feature.

## Pages

- [Super Artifacts](https://superartifacts.app/index.md): A hosting and data plane for artifacts an AI agent builds. The agent deploys a small web app to a live URL, people use it, and what they enter flows back to the agent over MCP. The platform runs no model; your agent writes every byte.
- [Deploy](https://superartifacts.app/features/deploy.md): One request turns a folder of files into a URL somebody can open on a phone. No build step, no bundler, no configuration.
- [Versions](https://superartifacts.app/features/versions.md): Publishing moves a pointer at an immutable version. Rolling back and rolling forward are the same operation.
- [Access](https://superartifacts.app/features/access.md): Artifacts are private by default. Sharing is an explicit grant, enforced by a signature the serving worker verifies.
- [Analytics](https://superartifacts.app/features/analytics.md): Whether anyone read an artifact, on what, and whether it threw an error while they did. The visitor address is never stored, only hashed.
- [Agents](https://superartifacts.app/features/agents.md): Install what the harness expects, one key from the dashboard, and the agent deploys on its own, including agents with no shell.
- [Pricing](https://superartifacts.app/pricing.md): Free to start. Solo is $19 a month or $190 a year, with a 14-day trial. Business is coming.
- [About](https://superartifacts.app/about.md): Why Super Artifacts exists, what it is built on, and an honest list of what is not finished.