Fully Local VMware Operations with Ollama

Run VMware infrastructure operations using a local LLM — no cloud API keys required.

Prerequisites

• Ollama installed: https://ollama.com
• vmware-aiops installed: uv tool install vmware-aiops
• VMware config ready: ~/.vmware-aiops/config.yaml + ~/.vmware-aiops/.env

Step 1: Pull a local model

# Recommended: Qwen2.5-Coder 32B (best tool-calling accuracy, needs 24GB VRAM)
ollama pull qwen2.5-coder:32b

# Alternative: 14B (needs 10GB VRAM)
ollama pull qwen2.5-coder:14b

# Lightweight: 7B (needs 6GB VRAM, lower accuracy for complex operations)
ollama pull qwen2.5-coder:7b

Step 2: Choose your agent

Option A: Aider (simplest)

# Install aider
pip install aider-chat

# Run with Ollama + vmware-aiops conventions
aider --conventions codex-skill/AGENTS.md --model ollama/qwen2.5-coder:32b

Then ask in natural language:

> List all VMs on my lab ESXi
> Show active alarms on vcenter-prod
> What's the datastore usage?

Option B: Goose (Block)

Edit ~/.config/goose/config.yaml:

extensions:
  vmware-aiops:
    name: VMware AIops
    cmd: vmware-aiops-mcp
    enabled: true
    type: stdio
    timeout: 300
    envs:
      VMWARE_AIOPS_CONFIG: ~/.vmware-aiops/config.yaml

Then:

goose session
> Check health status of all my VMware hosts

Option C: Continue (VS Code)

Add to your Continue MCP config:

mcpServers:
  - name: vmware-aiops
    command: vmware-aiops-mcp
    env:
      VMWARE_AIOPS_CONFIG: ~/.vmware-aiops/config.yaml

Configure Ollama as your model provider in Continue settings.

Step 3: Read-only mode (recommended for production)

For production environments, use vmware-monitor instead:

uv tool install vmware-monitor

# Aider
aider --conventions codex-skill/AGENTS.md --model ollama/qwen2.5-coder:32b

# Or configure MCP with vmware-monitor-mcp instead of vmware-aiops-mcp

vmware-monitor has zero destructive operations in its codebase — safe to use with any model.

Model comparison for VMware operations

Model	VRAM	Tool calling	Complex ops	Recommended for
Qwen2.5-Coder 32B	24GB	~90%	Good	Full operations
Qwen2.5-Coder 14B	10GB	~80%	Fair	Monitoring + simple ops
Qwen2.5-Coder 7B	6GB	~60%	Poor	Monitoring only
DeepSeek-Coder-V2 16B	12GB	~75%	Fair	Alternative to Qwen

Troubleshooting

# Verify vmware-aiops config and connection
vmware-aiops inventory vms --target <your-target>

# Verify Ollama is running
ollama list

# Check MCP server works standalone
echo '{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}},"id":1}' | vmware-aiops-mcp