# Deployment Guide
Deploy your MCP Content Credentials server to a permanent public URL.
## Option 1: Render.com (Recommended - Free Tier)
### Prerequisites
- GitHub account
- Render.com account (free)
### Steps
1. **Push your code to GitHub**
```bash
git add .
git commit -m "Ready for deployment"
git push origin main
```
2. **Connect to Render**
- Go to [https://render.com](https://render.com)
- Sign in with GitHub
- Click **"New +"** → **"Web Service"**
3. **Configure the deployment**
- Select your `mcp-content-credentials` repository
- Render will auto-detect the `render.yaml` configuration
- Click **"Apply"**
4. **Wait for deployment**
- First build takes ~5-10 minutes (installs c2patool + TrustMark)
- You'll get a URL like: `https://mcp-content-credentials.onrender.com`
5. **Update ChatGPT**
- Go to ChatGPT → Settings → Apps & Connectors → Your App → Edit
- Change MCP Server URL to: `https://your-app.onrender.com/mcp`
### Free Tier Limitations
- ⚠️ **Server spins down after 15 minutes of inactivity**
- First request after spin-down takes ~30-60 seconds
- 750 hours/month free (roughly 31 days)
### Keep server awake (optional)
Add a cron job to ping your server every 10 minutes:
- Use [cron-job.org](https://cron-job.org) (free)
- URL: `https://your-app.onrender.com/health`
- Schedule: Every 10 minutes
---
## Option 2: Railway (Recommended - $5/month)
### Prerequisites
- GitHub account
- Railway account
### Steps
1. **Install Railway CLI**
```bash
npm install -g @railway/cli
```
2. **Login and deploy**
```bash
railway login
railway init
railway up
```
3. **Add a domain**
```bash
railway domain
```
4. **Get your URL**
- Railway will provide: `https://your-app.up.railway.app`
- Use: `https://your-app.up.railway.app/mcp` in ChatGPT
### Pricing
- $5/month for hobby plan
- Always-on (no spin-down)
- Better performance than free tiers
---
## Option 3: Docker (Self-hosted or Cloud)
### Deploy with Docker
1. **Build the image**
```bash
docker build -t mcp-content-credentials .
```
2. **Run locally**
```bash
docker run -p 3001:3001 mcp-content-credentials
```
3. **Deploy to any cloud**
- **Google Cloud Run**: `gcloud run deploy`
- **AWS ECS**: Push to ECR and deploy
- **Azure Container Apps**: `az containerapp create`
- **DigitalOcean App Platform**: Connect GitHub repo
### Docker Hub (for easy deployment)
1. **Push to Docker Hub**
```bash
docker tag mcp-content-credentials yourusername/mcp-content-credentials
docker push yourusername/mcp-content-credentials
```
2. **Deploy anywhere that supports Docker**
---
## Option 4: Heroku
### Prerequisites
- Heroku account
- Heroku CLI installed
### Steps
1. **Login to Heroku**
```bash
heroku login
```
2. **Create app**
```bash
heroku create your-mcp-server
```
3. **Add buildpacks**
```bash
heroku buildpacks:add heroku/nodejs
heroku buildpacks:add heroku/python
```
4. **Deploy**
```bash
git push heroku main
```
5. **Your URL**
- `https://your-mcp-server.herokuapp.com/mcp`
### Pricing
- Free tier discontinued
- Starts at $7/month for Eco plan
---
## Environment Variables
All platforms support environment variables. Set these if needed:
```bash
NODE_ENV=production # Set to production
MCP_PORT=3001 # Port (usually auto-set by platform)
LOG_LEVEL=info # info, debug, warn, error
```
---
## Post-Deployment Checklist
### 1. Test your deployment
```bash
# Health check
curl https://your-url.com/health
# Should return:
# {"status":"ok","service":"mcp-content-credentials",...}
```
### 2. Test the MCP endpoint
```bash
curl -N -H "Accept: text/event-stream" https://your-url.com/mcp
```
### 3. Update ChatGPT App
- Go to ChatGPT → Settings → Apps & Connectors
- Edit your app
- Change MCP Server URL to: `https://your-url.com/mcp`
### 4. Test with ChatGPT
Ask: "Check this image: https://example.com/image.jpg"
---
## Troubleshooting
### Build fails - c2patool installation
**Render/Railway**: Should auto-install via `scripts/install-trustmark.cjs`
**Fix**: Check logs for Python or c2patool errors
```bash
# Render: View logs in dashboard
# Railway: railway logs
```
### Server starts but /mcp times out
- Check if port is correct (should be 3001)
- Verify health endpoint works: `/health`
- Check server logs for errors
### ChatGPT says "Server not found"
- Make sure URL ends with `/mcp`
- Verify server is running (check health endpoint)
- Try reconnecting the app in ChatGPT
### High latency / slow responses
- Free tiers have cold starts (30-60s on first request)
- Consider paid plans for always-on servers
- Use keep-alive cron jobs for free tiers
---
## Monitoring
### Render Dashboard
- View logs: Dashboard → Service → Logs
- Monitor uptime: Dashboard → Metrics
### Railway Dashboard
- Logs: `railway logs`
- Metrics: Railway dashboard
### Health Check Monitoring
Use **UptimeRobot** (free) to monitor your server:
1. Go to [uptimerobot.com](https://uptimerobot.com)
2. Add monitor: `https://your-url.com/health`
3. Get alerts if server goes down
---
## Cost Comparison
| Platform | Free Tier | Paid | Always-On | Cold Start |
|----------|-----------|------|-----------|------------|
| **Render** | ✅ 750h/mo | $7/mo | ❌ | 30-60s |
| **Railway** | ❌ | $5/mo | ✅ | No |
| **Heroku** | ❌ | $7/mo | ✅ | No |
| **Fly.io** | ✅ Limited | $5/mo | Depends | Varies |
**Recommendation**:
- **Testing/Personal**: Render free tier
- **Production**: Railway ($5/mo)
---
## Need Help?
- Check server logs first
- Test `/health` endpoint
- Verify ngrok still works locally (to rule out code issues)
- Check platform-specific documentation
## Security Notes
- Server only accepts URL-based verification (no file uploads)
- CORS enabled for all origins (consider restricting in production)
- No authentication by default (consider adding for production use)
- All verification is read-only
