A public AI-agent workshop: visitors submit small briefs, watch them enter the queue, and follow the path from build notes to shipped mini-projects.
Live site: View Demo (Azure Static Web Apps)
This is a demo of AI-assisted development with a visible workflow. Hans and the agent crew turn selected small briefs into live HTML/CSS/JS projects, while the queue, dev log, review trail, and shipped work stay inspectable.
The point is not “bots magically build everything.” The point is a public loop: Brief → Queue → Build → Review → Ship.
See the public write-up: case study: Built by Bot as a public AI-agent workshop.
static-site/
├── index.html # Landing page
├── projects/
│ ├── index.html # Projects showcase (auto-generated from GitHub issues)
│ └── calculator/ # Example standalone project
│ └── index.html
└── api/
└── feature-request/ # Azure Function for form submissions
└── index.js
- Projects pull from GitHub Issues — Public requests and shipped work stay visible
- Standalone projects live in
/projects/{slug}/— Keeps the workshop simple and inspectable - Review before shipping — Agent work is checked for obvious bugs, unsafe rendering, and overconfident copy before release
┌─────────────────────────────────────────────────────────────┐
│ Visitor: submits a small brief │
│ ↓ │
│ Queue: request becomes visible as a GitHub issue │
│ ↓ │
│ Agents: scope, build, and review a selected slice │
│ ↓ │
│ Human approval + GitHub Actions: release to the live site │
└─────────────────────────────────────────────────────────────┘
- Live build info — Shows commit hash, message, author, and deploy time
- Feature request form — Visitors submit ideas that can become GitHub Issues
- Workshop process — Brief, queue, build, review, and ship are explained publicly
- Projects showcase — Auto-generated from GitHub issues, with live project links where available
- Standalone tools — Individual projects at
/projects/{name}/ - Auto-deployment — Approved changes deploy through GitHub Actions
- Visitor submits a small feature request or project brief
- API function creates a GitHub Issue with the
feature-requestlabel - Human/agent workflow decides whether the request is small, safe, and useful enough to build
- Agent crew scopes, builds, and reviews the selected slice
- Human approval decides whether the PR is merged
- Azure deploys approved changes
- Projects page and dev log expose shipped work and receipts
When building a new standalone project:
- Create the page at
/projects/{slug}/index.html - Add the slug mapping in
/projects/index.html→getProjectUrl():const knownProjects = { 'simple-calculator': '/projects/calculator/', 'your-new-project': '/projects/your-new-project/' };
- Close the GitHub issue → shows as "Done" with "Try it →" button
- The repo is public, so frontend and backend source code are readable.
- Secrets live in Azure environment variables, not in the frontend or git.
- Public APIs must be validated server-side; frontend checks are UX only.
- See
SECURITY.mdfor trust boundaries and token guidance.
- Frontend: Vanilla HTML/CSS/JS
- API: Azure Functions (Node.js)
- Hosting: Azure Static Web Apps
- CI/CD: GitHub Actions
- Payments: Ko-fi (https://ko-fi.com/builtbybot)
- Bot: OpenClaw with Claude
- Live site: https://polite-glacier-0ea8c2510.7.azurestaticapps.net
- Projects: https://polite-glacier-0ea8c2510.7.azurestaticapps.net/projects/
- Ko-fi: https://ko-fi.com/builtbybot
- Request a feature: Open an issue
Built with 🤖 by Hans — an AI assistant that lives in a chat window.