Keep every artifact.
Open Artifacts keeps the pages, docs, code and PDFs from any AI agent (Claude Code, Codex, Cursor, OpenCode) on your own Cloudflare account.
Tell your agent:
set up open-artifacts.com/llms.txt
Set up in three steps #
-
Tell your agentYour agent reads llms.txt, checks the prerequisites, clones the latest release, and runs pnpm install.
-
Sign in to CloudflareOne step needs you:
npx wrangler login. Setup then creates D1, R2 and KV and the Access sign-in app, deploys to both domains, and adds the MCP server to your agent. -
Publish and shareYour agent publishes a Welcome artifact and gives you the link. Later, ask it to write up the Q4 data migration plan and share it with a teammate.
What you get #
-
HTML, Markdown, code and PDFInteractive pages, rendered docs, highlighted code and PDFs, viewed in a web dashboard with search.
-
Every version keptEach publish is a new version that records which agent made it. View or restore any of them.
-
Private until you shareShare with people by email as viewer or editor, or make a public link that search engines don't index.
-
Live pagesAn HTML artifact can save and read shared data, so polls, trackers and checklists just work.
-
Any MCP agentClaude Code, Codex CLI, Cursor, OpenCode, ChatGPT, or anything else that speaks MCP.
-
Yours to runMIT-licensed, on your own domain, and works on the Workers Free plan. Nothing passes through us.
Safe to open #
Agent-written HTML runs in a sandboxed frame on a separate, cookieless origin. It can't read your session, call the API as you, or reach the dashboard.
- Sandboxed
- Each artifact runs in its own opaque origin, so it can't see the dashboard or other artifacts.
- No passwords
- Sign-in goes through Cloudflare Access with a one-time PIN, GitHub, or Google.
- Private by default
- New artifacts are visible only to their creator. Anything you can't see returns a 404, so a link never reveals that it exists.
- Never indexed
- Public links send noindex headers, and robots.txt blocks crawlers on both hosts.
- Revocable tokens
- Agent tokens are stored only as hashes, and you can revoke them in settings.
Open source, MIT licensed #
Read the code, file an issue, or run it yourself.