π‘οΈ N2 Stitch MCP β Resilient Proxy Skill
Never lose a screen generation again. The only Stitch MCP proxy with TCP drop recovery.
The Problem
Google Stitch's generate_screen_from_text takes 2β10 minutes, but the API drops TCP after ~60 seconds.
Other MCP servers: Request β 60s β TCP drop β β LOST!
N2 Stitch MCP: Request β 60s β TCP drop β π‘οΈ Auto-recovery β β
Delivered!
Why This One?
| Feature | Others | N2 Stitch MCP |
|---|---|---|
| TCP Drop Recovery | β | β Auto-polling |
| Generation Tracking | β | β
generation_status |
| Exponential Backoff | β | β 3x retry + jitter |
| Auto Token Refresh | β οΈ | β Background refresh |
| Test Suite | β | β 35 tests |
Quick Setup
1. Authenticate (one-time)
# Option A: gcloud (recommended)
gcloud auth application-default login
# Option B: API Key
export STITCH_API_KEY="your-key"
2. Add to MCP Config
{
"mcpServers": {
"n2-stitch": {
"command": "npx",
"args": ["-y", "n2-stitch-mcp"]
}
}
}
Available Tools
Stitch API (auto-discovered)
- create_project β Create a Stitch project
- list_projects β List all projects
- get_project β Get project details
- list_screens β List screens in a project
- get_screen β Get screen HTML/CSS
- generate_screen_from_text β β¨ Generate UI from text (Resilient!)
- edit_screens β Edit existing screens
- generate_variants β Generate design variants
Virtual Tools (N2 Exclusive)
- generation_status β Check generation progress in real-time
- list_generations β List all tracked generations
Links
- NPM: https://www.npmjs.com/package/n2-stitch-mcp
- GitHub: https://github.com/choihyunsus/n2-stitch-mcp
- Website: https://nton2.com
Part of the N2 AI Body series β Building the Body for AI






