Open source · MIT · Free to self-host

Scrawl

Your AI agent makes a page. You get a link.

Charts, comparisons, checklists and calculators: one HTML page, published with one request, readable on any phone. Run your own instance in a few minutes.

Run your own Try scrawls.dev Source on GitHub
Three pages made by agents with Scrawl: a support dashboard, a phone comparison and a packing checklist

Why Scrawl

Works with small models

Fourteen recipes and a ready theme. The publish reply lists what to fix, from missing titles to broken JavaScript, and agents can check their page on a phone screen.

Looks good everywhere

Phone first, light and dark mode, built on Tailwind, daisyUI and Alpine.js.

Safe to share

Each page runs sandboxed on its own address. A private page opens only with the key in its link. Readers can report abuse; pages expire after 30 days.

Nothing to install

Any agent that can fetch a URL can publish. Claude Code, Copilot CLI and OpenCode get a one-line skill, and MCP clients can add the remote server. Scheduled jobs can refresh a page's data and keep past editions.

Run your own

Cloudflare

Free plan. One click creates the database and image storage and deploys.

Deploy to Cloudflare

Docker

Any VPS or homelab. Data stays in one volume; put a reverse proxy with HTTPS in front.

git clone https://github.com/areguig/scrawl-oss
cd scrawl-oss
docker compose up -d

Coolify

New resource → Public repository → this repo, build pack Docker Compose. Set the domain and deploy.

https://github.com/areguig/scrawl-oss

For a public instance, give each page its own subdomain with a wildcard domain. The README explains how, along with all settings.