WiseTech Image MCP Server
Enables image upscaling, face enhancement, and background removal via public image URLs, returning base64 PNG results.
README
WiseTech Image MCP Server
An MCP (Model Context Protocol) server that exposes all three WiseTech AI image tools to any MCP-compatible client (Claude, ChatGPT, Grok, etc.):
| Tool | What it does | API endpoint |
|---|---|---|
upscale_image |
Upscales an image to higher resolution | {main_api}/upscale_web_v2 |
enhance_face |
Enhances faces in an image | {main_api}/faceenhance_web |
remove_background |
Removes the background from an image | {bg_api}/uploadImageV2 → /processImageBackgroundRemove → /getStatus → /downloadImage |
All three accept a public image URL and return the processed image as base64 PNG.
Verified working — all three APIs tested end-to-end on 2026-06-20.
Quick start (local testing)
npm install
npm run dev
Server starts on http://localhost:3000/mcp. No .env needed for local testing — it auto-resolves the API URLs.
Environment variables
All config lives in .env (never committed to git). Copy .env.example to .env and fill in:
| Variable | Required | Description |
|---|---|---|
MCP_AUTH_KEY |
Yes for production | Your secret key. Clients send Authorization: Bearer <key>. Generate with openssl rand -hex 32. |
WISETECH_MAIN_DNS |
No | DNS resolver URL for main API (default: https://sd.rad-wi.com) |
WISETECH_BG_DNS |
No | DNS resolver URL for bg remover API (default: https://ibgc.rad-wi.com/) |
WISETECH_MAIN_FALLBACK |
No | Fallback main API IP if DNS fails |
WISETECH_BG_FALLBACK |
No | Fallback bg API IP if DNS fails |
WISETECH_MAIN_API_URL |
No | Pin a specific main API host (skips DNS) |
WISETECH_BG_API_URL |
No | Pin a specific bg API host (skips DNS) |
WISETECH_API_KEY |
No | API key if WiseTech adds one (currently not needed) |
WISETECH_AUTH_HEADER |
No | Header name for the API key (e.g. x-api-key) |
PORT |
No | HTTP port (default: 3000, use 8443 for VPS HTTPS) |
HTTPS_CERT_DIR |
No | Path to Let's Encrypt cert dir. When set, server uses HTTPS directly (no nginx needed). Example: /etc/letsencrypt/live/mcp.yourdomain.com |
DNS auto-refresh
The WiseTech API IPs rotate daily. This server replicates the WordPress plugin's DNS-resolution: on startup (and every 6 hours) it fetches the current URL from the DNS endpoints. No manual IP updates needed.
Deploy on Heroku
Step 1: Install Heroku CLI and login
npm install -g heroku
heroku login
Step 2: Create a Heroku app
cd wisetech-image-mcp
heroku create wisetech-image-mcp
Step 3: Set environment variables
# Generate a strong auth key
heroku config:set MCP_AUTH_KEY=$(openssl rand -hex 32)
# API URLs (auto-resolved from DNS, but set fallbacks)
heroku config:set WISETECH_MAIN_DNS=https://sd.rad-wi.com
heroku config:set WISETECH_BG_DNS=https://ibgc.rad-wi.com/
heroku config:set WISETECH_MAIN_FALLBACK=http://34.232.100.156:5454
heroku config:set WISETECH_BG_FALLBACK=http://35.190.164.209:5454
Step 4: Deploy
git add .
git commit -m "Initial deploy"
git push heroku main
Heroku automatically:
- Runs
npm install - Runs
npm run build(via thebuildscript inpackage.json) - Starts the app using the
Procfile:web: node dist/index.js - Assigns a port via the
PORTenv var (handled automatically)
Step 5: Verify it's running
heroku open
# Or test the MCP endpoint:
curl -X POST https://wisetech-image-mcp.herokuapp.com/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "Authorization: Bearer YOUR_MCP_AUTH_KEY" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
You should see all three tools in the response.
Step 6: Connect to AI tools
Your MCP URL is: https://wisetech-image-mcp.herokuapp.com/mcp
Use this URL + your MCP_AUTH_KEY in Claude / ChatGPT / Grok (see below).
Heroku notes
- Free tier sleeps after 30 min of inactivity. First request after sleep takes ~10s to wake up. Use
heroku ps:scale web=1and consider upgrading to Eco/Hobby for always-on. - Port: Heroku sets
PORTautomatically — don't set it yourself. - Timeout: Heroku has a 30s request timeout on free tier. Image processing can take up to 180s. If you hit timeouts, upgrade to Hobby tier or use a VPS instead.
Deploy on a VPS (recommended for production)
VPS is better than Heroku for this because image processing can take up to 3 minutes (Heroku free tier times out at 30s).
This guide is written for your specific VPS (Ubuntu 24.04, IP 167.88.43.163).
Your VPS current state
| Service | Port | Status |
|---|---|---|
| Docker (n8n) | 3000 | ⚠️ Taken — don't use |
| OpenVPN | 80 | ⚠️ Taken — don't use |
| nginx | 443, 8080 | ⚠️ Taken (n8n.conf) |
| Redis | 6379 | In use |
| SSH | 22 | In use |
| 8443 | free | ✅ We'll use this |
Strategy: Use port 8443 with HTTPS directly from Node (no nginx needed). This avoids touching any existing services — no conflicts with n8n, OpenVPN, or nginx.
Step 1: Install pm2 (process manager)
pm2 is not installed on your VPS yet. SSH into your VPS and install it:
ssh root@167.88.43.163
npm install -g pm2
Step 2: Clone the repo
If the repo is public:
cd /opt
git clone https://github.com/Motasaith/MCP_Upscale_Face-Enhancer_Remove-BG.git wisetech-mcp
cd wisetech-mcp
npm install
npm run build
If the repo is private, generate a Personal Access Token first:
- Go to https://github.com/settings/tokens → "Generate new token (classic)" → check
repo→ generate - Copy the token, then on the VPS:
cd /opt
git clone https://Motasaith:YOUR_TOKEN@github.com/Motasaith/MCP_Upscale_Face-Enhancer_Remove-BG.git wisetech-mcp
cd wisetech-mcp
npm install
npm run build
Note: The repo is safe to make public — all secrets are in
.env(gitignored),SECURITY.mdis gitignored, and the PHP reference folder is gitignored. The code on GitHub contains no API URLs or keys.
Step 3: Create the .env file
nano /opt/wisetech-mcp/.env
MCP_AUTH_KEY=GENERATE_A_STRONG_KEY_HERE
WISETECH_MAIN_DNS=https://sd.rad-wi.com
WISETECH_BG_DNS=https://ibgc.rad-wi.com/
WISETECH_MAIN_FALLBACK=http://34.232.100.156:5454
WISETECH_BG_FALLBACK=http://35.190.164.209:5454
WISETECH_API_KEY=
WISETECH_AUTH_HEADER=
PORT=8443
HTTPS_CERT_DIR=/etc/letsencrypt/live/mcp.yourdomain.com
Generate a strong auth key:
openssl rand -hex 32
Lock down the file:
chmod 600 /opt/wisetech-mcp/.env
chown root:root /opt/wisetech-mcp/.env
Step 4: Point a subdomain at this VPS
In your DNS dashboard (wherever you manage your domain — Cloudflare, Namecheap, etc.), add an A record:
A mcp.yourdomain.com → 167.88.43.163
Wait a few minutes for DNS to propagate. Verify:
dig mcp.yourdomain.com +short
# should return 167.88.43.163
Step 5: Get the SSL certificate (DNS challenge — no port 80 needed)
This method avoids touching port 80 (which OpenVPN uses). It uses a DNS TXT record instead.
apt install -y certbot
certbot certonly --manual --preferred-challenges dns -d mcp.yourdomain.com
Certbot will print something like:
Please deploy a DNS TXT record under:
_acme-challenge.mcp.yourdomain.com
with this value:
abc123def456...
Go to your DNS dashboard, add that TXT record, wait ~2 minutes, then press Enter in the terminal.
Certbot saves the cert to /etc/letsencrypt/live/mcp.yourdomain.com/.
Note: This cert expires in 90 days. Since it's a manual DNS challenge, you'll need to repeat this step before it expires. If your domain is on Cloudflare, you can automate it later with certbot-dns-cloudflare.
Step 6: Start the server with pm2
cd /opt/wisetech-mcp
pm2 start dist/index.js --name wisetech-mcp
pm2 save
pm2 startup # follow the instructions it prints to make it survive reboots
Check it's running:
pm2 status
pm2 logs wisetech-mcp --lines 10
You should see:
WiseTech Image MCP server listening on :8443 (HTTPS, POST /mcp)
Tools: upscale_image, enhance_face, remove_background
[DNS] APIs resolved successfully.
Step 7: Open the port (if firewall is active)
ufw status
If it says "active":
ufw allow 8443/tcp
If it says "inactive", skip this step.
Step 8: Test from outside
curl -i -X POST https://mcp.yourdomain.com:8443/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "Authorization: Bearer YOUR_MCP_AUTH_KEY" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
You should see all three tools in the response.
Step 9: Connect to AI tools
Your MCP URL is: https://mcp.yourdomain.com:8443/mcp
Use this in Claude / ChatGPT / Grok (see the "Connecting to AI tools" section above).
Useful pm2 commands
pm2 status # see if it's running
pm2 logs wisetech-mcp # live logs (Ctrl+C to exit)
pm2 restart wisetech-mcp # restart after code changes
pm2 stop wisetech-mcp # stop
pm2 delete wisetech-mcp # remove from pm2
Certificate renewal (every 90 days)
Since we used a manual DNS challenge, auto-renewal won't work unattended. Before the cert expires:
certbot certonly --manual --preferred-challenges dns -d mcp.yourdomain.com
pm2 restart wisetech-mcp
If your domain is on Cloudflare, you can automate this:
apt install -y python3-certbot-dns-cloudflare
# Configure with your Cloudflare API token, then:
certbot certonly --dns-cloudflare --dns-cloudflare-credentials /root/.cloudflare.ini -d mcp.yourdomain.com
# Then add a cron job for auto-renewal
Connecting to AI tools
Claude Desktop
Edit this file:
- Windows:
%APPDATA%\Claude\claude_desktop_config.json - Mac:
~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"wisetech-image": {
"url": "https://your-domain.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_MCP_AUTH_KEY"
}
}
}
}
Restart Claude Desktop. You'll see three tools available:
upscale_imageenhance_faceremove_background
ChatGPT / Grok / other MCP clients
Go to their custom tools / MCP settings and add:
- URL:
https://your-domain.com/mcp - Header:
Authorization: Bearer YOUR_MCP_AUTH_KEY
Local testing (no deployment needed)
npm run dev
Point your AI client at http://localhost:3000/mcp (no auth key needed for local).
How it works
Upscale & Face Enhance (synchronous)
User → AI tool → MCP server → downloads image from URL → base64-encodes →
POST {main_api}/upscale_web_v2 (or /faceenhance_web)
body: {"image": "<base64>"}
→ response: {"image": "<base64 result>"}
→ returns base64 PNG to the AI tool → AI tool shows it to the user
Background Remover (asynchronous, handled automatically)
User → AI tool → MCP server → downloads image →
1. POST {bg_api}/uploadImageV2 (multipart "file")
2. GET {bg_api}/processImageBackgroundRemove?imageName=...
3. Poll {bg_api}/getStatus?imageName=... (every 3s until status=1)
4. GET {bg_api}/downloadImage?imageName=...png
→ returns base64 PNG to the AI tool
The entire async flow is hidden from the AI tool — it just gets the final image back.
Supported formats
| Tool | Formats |
|---|---|
| Upscale | JPG, JPEG, PNG, GIF, JFIF, WEBP, BMP, ICO, SVG, AVIF |
| Face Enhance | JPG, JPEG, PNG, GIF, JFIF, WEBP, BMP, ICO, SVG, AVIF |
| Background Remove | JPG, JPEG, PNG, WEBP, JFIF, BMP |
Max file size: 10 MB (enforced server-side, matching the WP plugin).
Project structure
wisetech-image-mcp/
├── index.ts # MCP server — all 3 tools (supports HTTP + HTTPS)
├── package.json # Dependencies + scripts
├── tsconfig.json # TypeScript config
├── Procfile # Heroku deployment config
├── .env # Your secrets (NEVER commit to git)
├── .env.example # Template for .env
├── .gitignore # Blocks .env, SECURITY.md, node_modules from git
├── SECURITY.md # Security & deployment details (kept private)
├── README.md # This file
└── Extra_Just_For_API_Info/ # Original PHP plugin reference (gitignored)
Notes
- The server uses Streamable HTTP transport (not stdio), so it works with remote/web-based MCP clients like ChatGPT and Grok, not just Claude Desktop.
- Background removal output is always PNG (the API converts the input filename to
.png). - All API URLs are in
.env, not in the code — someone reading the source sees nothing. - See
SECURITY.mdfor detailed security information and hardening steps.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.