lukataylo

Foldo — AI skill for Claude Code

AI community

A Figma-style multiplayer review canvas for AI-generated code.

How to install Foldo

This entry records only its repository, not the path inside it, so there is no exact command to give. Open lukataylo/foldo and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Foldo does

A Figma-style multiplayer review canvas for AI-generated code. Comment, compare, and ship changes Claude Code wrote — without leaving the canvas.

Alternatives in AI

  • Open Pencil — AI-native design editor 8k ★
  • Cursor Talk To Figma MCP — TalkToFigma: MCP integration between AI Agent (Cursor, Claude Code, Codex) and Figma, allowing Agentic AI to c 7k ★
  • Om Humanize — Voice-calibrated editing — makes Claude-drafted text sound like you wrote it, not like AI wrote it 4.6k ★

README

Foldo

Vibe-coding platform for hackathons: a team describes an app, an LLM writes it, and it runs live in the browser ([WebContainers](https://webcontainers.io)). Accounts, saved projects, public share links, a gallery with likes, and an organizer console. Based on [bolt.new](https://github.com/stackblitz/bolt.new) (MIT), ported to Node.

Run locally (Node >= 22.13)

cp .env.example .env     # add at least one model key; npm scripts load .env
npm install && npm run dev
npm run build && npm start      # production mode, port $PORT (default 3000)

Deploy to Railway

  1. New service from this repo (Dockerfile via railway.json, health check /healthz).
  2. Add a Volume at /data. SQLite lives there, so run exactly one instance.
  3. Set APP_SECRET, ADMIN_EMAILS, DB_PATH=/data/foldo.db, and one model key (or add keys later in /admin).
  4. Sign up with an ADMIN_EMAILS address, open /admin.

Running a hackathon

Everything is in **/admin**: paste provider keys (OpenRouter, DeepSeek, OpenAI, Xiaomi MiMo, Qwen, Kimi, GLM, MiniMax, or any OpenAI-compatible endpoint), press **Test**, pick the default, set an invite code, event name + countdown, an announcement banner, per-team and global daily message caps, and manage teams (disable, reset password) and the gallery. The base URLs and model ids for the Chinese providers are defaults to verify against each provider's docs; edit them in the admin card if they differ. Teams can pick between enabled models in the prompt box.

Safeguards: per-team daily quota, a global daily cap, one generation at a time per team, conversation size caps, failed generations are refunded, login/sign-up/invite-code throttling, Origin checks on writes, encrypted provider keys.

Event-night controls (/admin)

  • Live: builds running, requests per minute, failovers, errors and answers per model, top teams, estimated spend.
  • Emergency pause: one switch plus a message teams see; running builds finish, sa