Skip to main content

Introduction

vibeD is a workload orchestrator for GenAI-generated artifacts. It bridges vibe-coding tools (Claude, Gemini, ChatGPT) with your own Kubernetes infrastructure, letting AI-generated websites and web apps deploy directly to your cluster.

Why vibeD?​

Today's AI coding tools create amazing artifacts but run them in sandboxed environments. This creates problems for enterprises:

  • Data exposure - Confidential code runs on third-party infrastructure
  • No persistence - Artifacts disappear when the session ends
  • No integration - Can't connect to internal services or databases

vibeD solves this by providing an MCP server that AI coding tools can call to deploy artifacts to your own infrastructure.

How It Works​

AI Coding Tool (Claude, Gemini, etc.)
        |
        | MCP Protocol (deploy_artifact)
        v
    ┌─────────┐
    │  vibeD   │  ← MCP Server + Dashboard
    └────┬────┘
         |
    ┌────┴────┐
    │ Build   │  ← Buildah (in-cluster Jobs)
    └────┬────┘
         |
    ┌────┴──────────────┐
    │ Deploy to:        │
    │ • Knative Serving │  ← Preferred (serverless)
    │ • Kubernetes      │  ← Fallback (always available)
    │ • wasmCloud       │  ← WebAssembly (opt-in)
    └───────────────────┘

Key Features​

  • MCP Server - 7 tools for deploying, managing, and monitoring artifacts
  • Auto-detection - Automatically picks the best deployment target
  • Buildah builder - Auto-generates Dockerfiles, builds container images in-cluster via Kubernetes Jobs (no Docker socket required)
  • Dashboard - Web UI showing all deployed artifacts with status and URLs
  • Multiple storage backends - Local filesystem or GitHub repository
  • Container registry - Push/pull images to any OCI-compatible registry
  • Helm charts - Production-ready deployment for vibeD and its dependencies

Quick Start​

# Clone and build
git clone https://github.com/vibed-project/vibeD.git
cd vibed
make build

# Start with HTTP transport (dashboard + MCP)
./bin/vibed --config vibed.yaml --transport http

Then open http://localhost:8080 to see the dashboard.