vibegui/awesome-ai-native-cloudflare-app

AI-native apps on Cloudflare Workers: one Worker = HTTP app + MCP server + MCP-App UI (deco studio) + WhatsApp bot. Includes the master prompt.

About awesome-ai-native-cloudflare-app

A batteries-included starting point for AI-native apps on Cloudflare Workers: one Worker that is simultaneously an HTTP app, an MCP server (so AI agents and workspaces like deco studio can operate it), an embeddable MCP-App UI, and a WhatsApp bot — deployed by git push via Workers Builds.

Start with PROMPT.md — the master prompt / field guide. Read it yourself, or paste it into your AI coding agent (as CLAUDE.md or system context) and describe the app you want. The scaffold in this repo is the reference implementation of every pattern it describes.

The demo domain is a trivial notes app + WhatsApp assistant — deliberately boring, so the architecture is the interesting part. Replace the tools and views with your own.

The app measures itself (first-party analytics in D1 — no GA/PostHog: a track() helper, a POST /e beacon, cookieless daily-hash uniques, and the dimensions Cloudflare's own dashboard shows — status, cache, browser/OS/device, ASN, colo, IP range, referrer — on a cross-filtering analyticsoverview screen) and exposes goals, memory, and hypotheses as MCP tools. Connect Claude Code to the deployed app (.mcp.json — set APPMCPURL / APPMCPTOKEN), and CLAUDE.md turns any session into an improvement cycle:

Set a few goals (goalset), throw an agent at it, spend tokens, watch the metrics move. Humans stay in the loop at consequence: user-facing messages, money, data deletion. Full pattern in PROMPT.md §10.

From the project README.

Maintaining this server?

Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:

[![On MCP Radar](https://mcp.liqiwa.com/badge.svg)](https://mcp.liqiwa.com/s/vibegui--awesome-ai-native-cloudflare-app.html)

Related MCP servers