There's a certain type of developer who enjoys the ritual of SSH-ing into a fresh VPS, configuring systemd services, and setting up certbot. And then there's everyone else — people who want their AI assistant running in the cloud without spending an afternoon on infrastructure.
Railway and Cloudron are for the second group. Both give you a working ZeroClaw deployment without touching a terminal, and both are worth understanding because they suit different situations.
Railway: The Fastest Path to Cloud
Railway is a cloud platform that deploys from GitHub repos or Docker images. It handles provisioning, networking, and scaling automatically. ZeroClaw has official one-click templates that get you from zero to running in about two minutes.
Visit the ZeroClaw Railway template and click "Deploy Now." Railway provisions a container, pulls the ZeroClaw Docker image, sets up persistent storage for your memory database, and gives you a private URL for health checks. You don't write a single line of infrastructure config.
After deployment, set your configuration through environment variables in the Railway dashboard — Railway's preferred approach for secrets:
``` ZEROCLAW_AI_PROVIDER=anthropic ZEROCLAW_AI_MODEL=claude-sonnet-4-20250514 ZEROCLAW_AI_API_KEY=sk-ant-... ZEROCLAW_TELEGRAM_TOKEN=123456:ABC... ZEROCLAW_TELEGRAM_ALLOWED_USERS=123456789 ```
ZeroClaw reads environment variables as an alternative to config.toml, which makes it a natural fit for cloud deployments where you don't want secrets stored in files.
For persistent memory, mount a Railway volume at `/data` and set `ZEROCLAW_MEMORY_PATH=/data/memory.db`. Your conversation history then survives redeployments and container restarts.
The cost is where ZeroClaw's small footprint really pays off on Railway. Railway's usage-based pricing bills you for actual resource consumption. ZeroClaw uses ~4MB of RAM and near-zero CPU when idle. A personal assistant handling around 500 messages per day typically costs $2-5/month. Running OpenClaw on Railway costs $15-25/month due to its 1GB+ RAM usage — the same workload, five times the price.
Cloudron: Self-Hosting Without the Headaches
Cloudron is a different kind of platform. Instead of a managed cloud service, it's software you install on your own VPS that turns it into a personal app store. You get automatic updates, built-in backups, SSL certificate management, and a web UI for managing everything — without having to configure any of it yourself.
The appeal for ZeroClaw specifically is the operational automation. Cloudron pulls new ZeroClaw versions and updates them seamlessly. Your memory database is included in Cloudron's backup rotation automatically. SSL certificates are handled via Let's Encrypt without any configuration. And if you want to expose ZeroClaw's web gateway, Cloudron's auth system can protect it.
Setup starts with installing Cloudron on a VPS:
```bash wget https://cloudron.io/cloudron-setup chmod +x cloudron-setup ./cloudron-setup ```
After that, open the Cloudron dashboard, go to the App Store, search for ZeroClaw, and click Install. Configure your AI provider and channels through the settings UI. The whole process takes about 15 minutes.
Cost-wise, you need a VPS ($5-10/month for 1GB RAM, which is more than enough for ZeroClaw) plus Cloudron itself (free for up to 2 apps, $15/month for unlimited). Total: $5-25/month with full self-hosting control and none of the operational burden.
Choosing Between Them
| Feature | Railway | Cloudron | Bare VPS | |---------|---------|----------|----------| | Setup time | 2 minutes | 15 minutes | 30 minutes | | Auto-updates | On redeploy | Automatic | Manual | | Backups | Manual | Automatic | Manual | | Cost/month | $2-5 | $5-25 | $5-10 | | Control | Limited | Full | Full | | Best for | Quick deploy, testing | Self-hosters, teams | Power users |
Railway wins on speed and simplicity. Cloudron wins on automation and control. A bare VPS wins if you want maximum flexibility and don't mind managing things yourself.
Docker Deployment Anywhere
Both Railway and Cloudron use Docker under the hood. If you want to deploy ZeroClaw on any Docker-compatible platform — DigitalOcean App Platform, Fly.io, your own Kubernetes cluster — the same image works everywhere:
```bash git clone https://github.com/zeroclaw-labs/zeroclaw.git cd zeroclaw ./bootstrap.sh --docker ```
The Docker image is minimal — based on a distroless base image containing only the ZeroClaw binary and required certificates. Total image size: ~20MB. For comparison, a typical Node.js Docker image is 500MB+.
After Deployment
Regardless of which platform you choose, run through the same verification steps: confirm your bot responds on all configured channels, check the logs for connection errors, send a test message and restart the container to verify memory persistence, and set up monitoring through whatever the platform provides.
One thing worth noting about scaling: ZeroClaw handles hundreds of concurrent users on a single instance. If you genuinely need to scale beyond that, the bottleneck is almost always your AI provider's rate limits, not ZeroClaw itself. The 4MB footprint means you can run many instances on minimal hardware before infrastructure becomes a concern.