vibops-mcp

The provider-agnostic MCP server for GPU infrastructure — one interface for any cloud, any cluster, any provider.

The problem

Large enterprises and CSPs managing GPU infrastructure deal with fragmentation — AWS, GCP, Azure, on-prem, neoclouds, each with their own API, dashboard, and cost model. Correlating utilisation, cost, and workload type across providers requires jumping between 5 tools.

The solution

vibops-mcp is a single MCP server that abstracts this complexity. One pip install, 26 tools, and your AI assistant can observe, operate, and optimize your entire GPU fleet — regardless of where it runs.

• Observe — GPU utilisation, workload breakdown, MTTR, cost estimates, live K8s deployments
• Act — deploy models, scale clusters, run Helm/kubectl, trigger pipelines
• Configure — set cost rates, manage gateways, store secrets

All operations go through your VibOps instance and are recorded in the audit log.

Installation

pip install git+https://github.com/VibOpsai/vibops-mcp.git

Configuration

You need two environment variables:

Claude Desktop

Add to ~/.config/claude/claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "vibops": {
      "command": "vibops-mcp",
      "env": {
        "VIBOPS_URL": "https://vibops.example.com",
        "VIBOPS_TOKEN": "your-token-here"
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project root, or to the global config:

{
  "mcpServers": {
    "vibops": {
      "command": "vibops-mcp",
      "env": {
        "VIBOPS_URL": "https://vibops.example.com",
        "VIBOPS_TOKEN": "your-token-here"
      }
    }
  }
}

Claude Code (CLI)

claude mcp add vibops vibops-mcp \
  -e VIBOPS_URL=https://vibops.example.com \
  -e VIBOPS_TOKEN=your-token-here

Available tools

Observation (read-only)

Actions (write)

Configuration

Example prompts

"What's our GPU utilisation trend over the last 7 days?"
"Show me the cost breakdown per cluster this week."
"Deploy llama3:8b on vibops-dev with 2 replicas."
"Which clusters have open critical GPU alerts?"
"Scale the inference namespace to 4 replicas on prod-cluster."
"What's our MTTR for critical alerts?"

Roadmap

Wave 1 — Current (v0.1)

• ✅ Kubernetes (generic — on-prem, Kind, kubeadm)
• ✅ Any VibOps-connected cluster via gateway

Wave 2 — Q2 2026

• 🔜 NVIDIA DGX Cloud (NGC API)
• 🔜 AWS (EC2 GPU + EKS)
• 🔜 GCP (A3/A2 + GKE)
• 🔜 Azure (ND/NC series + AKS)

Wave 3 — Q3 2026

• 🔜 CoreWeave, Lambda Labs, RunPod
• 🔜 OVHcloud, Scaleway
• 🔜 Dell APEX, HPE GreenLake

Want a provider prioritised? Open an issue.

Contributing

See CONTRIBUTING.md. All contributions require a DCO sign-off (git commit -s).

License

MIT — free to use, modify, and distribute. See LICENSE.

Built on FastMCP and the VibOps platform.