Vibe Deploy: Claude Code to Live App in Minutes

You know that feeling when you're in the zone with Claude Code, your app is coming together, and then you hit the wall? The deployment wall. Suddenly you're wrestling with servers, databases, configs, and all the stuff that has nothing to do with your actual idea.

That's over now.

Vibe Deploy lets you go straight from coding with AI to a live, running application. Just tell your AI to deploy to RunOS and it handles everything: provisioning databases, setting up services, building your app, and getting it online. Under 10 minutes from idea to live URL.

Getting started

Three steps. That's it.

1. Create a RunOS account at console.beta.runos.com
2. Open the Vibe Deploy wizard from the cluster selector → Vibe Deploy and follow the instructions to install the RunOS CLI
3. Configure your project by running this in your project directory:

runos mcp configure claude

That command automatically sets up the MCP servers that connect your AI to RunOS. Once it's configured, your AI can provision services and deploy your app directly.

And here's the best part: you don't need Git, Docker, or any runtime installed on your machine. No Go, no Node, no Python. Everything builds and runs on your RunOS environment. Your local machine just needs Claude Code and the RunOS CLI.

What can it provision?

Right now, the MCP server supports:

• PostgreSQL for relational databases
• MySQL for relational databases
• Valkey (Redis-compatible) for caching and key-value storage
• MinIO for S3-compatible object storage

More services are coming. But these four cover the vast majority of what you need to build real applications.

See it in action

Here are some real prompts you can paste into Claude Code right now. Each one builds a complete, working application and deploys it to RunOS. All in under 10 minutes.

Talk to your services

This is where it gets really cool. The MCP connection isn't just for deploying. It's a full dev environment.

Every service you provision is accessible directly from your AI session. No need to open separate clients, remember connection strings, or switch between tools. Just ask.

Query your database:

• "How many comments do I have for each blog post?"
• "Show me all users who signed up this week"
• "Create a new database with a user that only has access to that database"

Manage your cache:

• "What's currently in my Valkey cache?"
• "Clear the cache"

Check your object storage:

• "How many files are in my blog-images bucket?"

Your AI talks to the service, gets the results, and answers in plain English. You stay in the flow. No context switching. No googling CLI commands you'll forget by tomorrow.

Faster debugging, for real

Here's why this matters more than convenience.

When something breaks, you normally jump between five different tools: your editor, the terminal, a database client, log viewers, your browser. By the time you piece together what went wrong, you've lost 20 minutes just gathering context.

With the RunOS MCP tool, your AI has access to all the relevant parts of your application at once: your code, app logs, build logs, and every service your app depends on. It can read the error in your app logs, check if the database migration ran correctly, look at what's in the cache, and trace the issue end to end.

Instead of you being the detective, your AI does the detective work. You just ask "why is my app returning 500 errors?" and it goes and finds out.

Security you can actually trust

Giving an AI access to your infrastructure sounds scary. We get it. That's why we split the RunOS MCP into four separate servers, each with a different permission level:

• Read: Gets information about your environment. Things like app status, logs, service health. Safe stuff
• Read sensitive: Reads information that includes sensitive data, like database passwords or API keys
• Write: Makes changes to your environment. Deploying apps, creating services, updating configs
• Write sensitive: Makes changes that involve sending or receiving sensitive information through the LLM

By default, only the read server runs without asking you first. Everything else prompts you before it executes, and clearly tells you which category the action belongs to. So you always know exactly what's about to happen and can decide if you're comfortable with it.

You stay in control. Your AI gets the access it needs to be useful, but never more than you've approved.

Want to lock it down further? You can remove any MCP server you don't want. Keep only the read server if you're extra cautious. Or remove just the two sensitive servers to make sure no passwords or API keys ever flow through the LLM. Mix and match to your comfort level.

And if there's something you'd rather not pass through the LLM at all? Just use the RunOS CLI or the web console directly. They have full access to everything the MCP servers can do. Your call.

Custom domains

Once your app is live, you probably want it on your own domain. Just ask your AI.

Seriously. Ask Claude Code something like "How do I get this app under my own domain?" and it will tell you exactly which IP address to point your DNS to. Update your domain's A record, and you're live on your own domain. It's that simple.

From prototype to production

Here's the thing most people don't realize: Vibe Deploy isn't just for quick demos.

What starts as a vibe coding session can easily graduate into a real production app. The same RunOS workflow scales with you:

• Start small: Vibe code your idea and deploy it on a single server. Perfect for proofs of concept, MVPs, or side projects
• Grow when ready: Add more servers for full redundancy and scaling when your app gets traction
• No migration needed: You don't jump to a different platform or rewrite anything. It's the same RunOS workflow from day one to day one thousand

You can also spin up multiple clusters with ease. A dev cluster, a staging cluster, and a production cluster. Each one is completely isolated, so you can test your changes properly before showing them to the public. Break things in dev all day long. Staging catches what you missed. Production stays clean.

Most tools make you choose upfront: toy project or serious infrastructure. RunOS doesn't. You start wherever you are and grow at your own pace, without jumping through hoops or switching platforms.

Why this matters

Deployment has always been the gap between having an idea and having a product. You can vibe code an entire app in 20 minutes, but then spend hours figuring out how to actually get it online. Servers, Docker, databases, DNS, SSL certificates, environment variables... it adds up fast.

Vibe Deploy closes that gap completely. Your AI already knows how to write code. Now it knows how to deploy it too.

The best part? This is just the beginning. We're adding more services, more integrations, and more AI tools. But right now, today, you can go from an idea to a live app with a database, object storage, and a custom domain in under 10 minutes.

Stop deploying. Start vibing.