🚀 Building Stream-Bridge: My Experience with Vibe Coding and AI Tools
Hello everyone! 👋
I recently completed a small project called Stream-Bridge — a peer-to-peer file transfer application built using WebSocket. What makes this project special is that I developed almost the entire thing using vibe coding with Claude Code.
In this post, I’ll walk you through how I built it, the challenges I faced, and what I learned along the way.
💡 Why I Built It
I’ve always been curious about how P2P file transfer works, and I wanted to build something lightweight that didn’t require third-party file hosting. I also wanted to explore how much I could push the limits of AI-assisted development using tools like Claude Code, Gemini CLI, and ChatGPT.
I tested Gemini CLI, but I found that Claude Code Pro offered better handling of large contexts and a smoother development experience.
📋 The Process: From Idea to Deployment
Before writing any code, I broke the entire process down into three key stages:
- Planning
- Development
- Deployment
1. 🧠Planning
Even though Stream-Bridge is a small project, planning was crucial. When using vibe coding with AI tools like Claude, clear direction matters a lot.
I used ChatGPT to brainstorm ideas, decide the tech stack (Node.js + WebSocket), and design a basic architecture.
Once the plan was ready, I wrote a detailed README.md and — more importantly — a Claude.md file. This file is used by Claude Code to retain context about your app, which makes future prompts more effective. I highly recommend it if you're using Claude!
2. 💻 Development
With the planning done, I started coding with Claude. I gave clear, structured prompts, and the tool generated about 97% of the application. It also helped fix bugs, improve structure, and made smart decisions without needing constant back-and-forth.
Features:
- Simple frontend for selecting and sending files
- Backend Node.js WebSocket server to handle the P2P transfer
The entire app was ready in 2 days.
3. 🚀 Deployment
Originally, I wanted to deploy the app on Vercel — it’s free and super easy. But I ran into a big issue: Vercel doesn’t support WebSocket, since it uses serverless architecture.
Claude didn’t mention this while generating the code, and by default, created a REST API-based version. This taught me an important lesson:
🧠Be very specific with your prompts when using AI tools.
To solve this, I looked into alternatives using ChatGPT. Some external WebSocket services were suggested, but they didn’t feel right.
So I set up a Node.js WebSocket server on a DigitalOcean droplet (similar to AWS EC2). This gave me full control and worked perfectly.
After updating the setup, I shared everything with Claude again, and it adjusted the app accordingly.
✅ Live demo: https://share.rsmk.dev
📂 GitHub Repo: https://github.com/manikandan97/stream-bridge
🤖 Final Thoughts: Will AI Replace Us?
No, AI won’t replace developers. But it will boost our productivity and let us focus more on ideas and less on boilerplate.
There’s still a lot of human work involved — setting up servers, debugging, deploying, testing edge cases, etc. AI is a tool, not a replacement.
Keep learning, keep building. The real risk is only if we stop learning.
Thanks for reading! 🙌 Feel free to check out the code and share your thoughts.
#AI #WebSocket #NodeJS #ClaudeCode #ChatGPT #DigitalOcean #BuildInPublic #streambridge #OpenSource #DeveloperJourney