vibops-mcp

vibops-mcp

The MCP server for VibOps — The AI Infrastructure Engine. From code to GPU in one conversation.

The problem

Getting an AI app from code to production on GPUs requires stitching together 9+ tools — git, Docker, CI/CD, Helm, kubectl, GPU monitoring, cost management, compliance, alerting. Each with its own API, dashboard, and cost model. No single interface spans the full pipeline.

The solution

vibops-mcp connects your AI assistant to VibOps — the engine that clones, builds, deploys, scales, monitors, fixes, and bills your apps and agents on any GPU, any cluster, any cloud. One pip install, 117 tools, one conversation.

• Ship — clone repos, build containers, deploy models, run Helm/kubectl, trigger pipelines, submit Slurm jobs
• Operate — scale deployments, manage VMs (Proxmox/XO/vSphere/HPE VME), detect and remediate GPU anomalies
• Observe — GPU utilisation, workload breakdown, MTTR, cost estimates, live K8s deployments
• Govern — AI Act compliance, SOC 2/RGPD reports, immutable audit chain, policy management
• FinOps — per-agent cost tracking, budget enforcement, chargeback, spend trends, waste analysis

Every operation goes through your VibOps instance and is recorded in the immutable audit log.

Installation

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

Configuration

You need two environment variables:

Variable	Description
VIBOPS_URL	Base URL of your VibOps instance, e.g. https://vibops.example.com
VIBOPS_TOKEN	API token — create one in VibOps → Settings → API Tokens

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 (16 tools — read-only)

Tool	Description
list_clusters	List clusters and GPU utilisation
list_kubectl_contexts	List available kubectl contexts
get_cluster_deployments	Live K8s deployment status for a cluster
get_cluster_rate	Get configured GPU cost rate for a cluster
list_jobs	List recent jobs with optional filters
get_job	Get job details and result
get_job_metrics	Job success rate, latency P50/P95/P99, error breakdown
get_gpu_metrics	Hourly GPU utilisation time-series
get_workload_breakdown	Job count by workload type
get_mttr	Mean Time To Resolve GPU alerts
get_cost_estimate	Estimated GPU spend
list_gateways	List registered gateways and status
list_alerts	List GPU alerts (open or resolved)
list_secrets	List secrets (names only, never values)
list_providers	List configured AI/GPU cloud providers
list_pipelines	List automation pipelines

Actions (18 tools — write)

Tool	Description
scale_deployment	Scale a K8s deployment replica count
deploy_model	Deploy an AI model onto a GPU cluster
helm_upgrade	Run helm upgrade --install (destructive — requires confirmed=True)
helm_uninstall	Uninstall a Helm release (destructive — requires confirmed=True)
run_kubectl	Run an arbitrary kubectl command
git_clone	Clone a git repository
create_secret	Store an encrypted secret
trigger_pipeline	Manually trigger an automation pipeline
slurm_get_cluster_info	Get Slurm cluster info and partition details
slurm_list_jobs	List Slurm jobs with optional filters
slurm_get_job_status	Get status of a specific Slurm job
slurm_get_job_output	Retrieve stdout/stderr of a completed Slurm job
slurm_submit_job	Submit a new Slurm job
slurm_cancel_job	Cancel a running or pending Slurm job
registry_list_repos	List container registry repositories
registry_list_tags	List tags for a container image
registry_check_image	Check image details (size, layers, created date)
registry_delete_tag	Delete a stale image tag (requires confirmed=True)

HPE VME / Morpheus (13 tools)

Tool	Description
vme_list_instances	List all VMs managed by HPE VME
vme_get_instance	Get detailed VM status and configuration
vme_list_servers	List physical hosts with CPU/memory
vme_list_clouds	List configured clouds/zones (KVM, VMware)
vme_start_instance	Start a stopped instance
vme_stop_instance	Stop a running instance
vme_restart_instance	Restart an instance
vme_create_snapshot	Create a snapshot (pre-migration)
vme_list_snapshots	List snapshots for an instance
vme_convert_image	Convert disk format (vmdk → qcow2) — VMware migration
vme_list_virtual_images	List available images and templates
vme_detect_vm_waste	Detect stopped/idle VMs with recommendations
vme_get_activity	Recent audit log from Morpheus

Hypervisors — Proxmox, vSphere, Xen Orchestra (30 tools)

Documented last and listed here in full: these were registered and served but absent from this README, so a reader counting the sections found 87 of the 117 the first line promises. They now add up.

Proxmox VE (5 tools)

Tool	Description
proxmox_list_vms	List VMs across the cluster with status and sizing
proxmox_start_vm	Start a stopped VM
proxmox_stop_vm	Stop a running VM (destructive — requires confirmed=True)
proxmox_migrate_vm	Migrate a VM to another node (destructive — requires confirmed=True)
proxmox_create_snapshot	Snapshot a VM before a risky change

VMware vSphere (9 tools)

Tool	Description
vsphere_list_vms	List VMs with power state, host and resources
vsphere_get_vm	Read one VM: power state, host, vCPU, RAM, NICs
vsphere_get_vm_metrics	CPU, memory and disk usage for one VM
vsphere_list_hosts	List ESXi hosts with capacity
vsphere_start_vm	Power a VM on
vsphere_stop_vm	Shut a VM down (destructive — requires confirmed=True)
vsphere_restart_vm	Restart a VM (destructive — requires confirmed=True)
vsphere_migrate_vm	vMotion a VM to another host (destructive — requires confirmed=True)
vsphere_create_snapshot	Snapshot a VM

Xen Orchestra / XCP-ng (15 tools)

Tool	Description
xo_list_vms	List VMs across all pools
xo_start_vm	Start a VM
xo_stop_vm	Cleanly shut a VM down (destructive — requires confirmed=True)
xo_migrate_vm	Migrate a VM to another host (destructive — requires confirmed=True)
xo_snapshot_vm	Snapshot a VM
xo_list_srs	List storage repositories — capacity planning before a migration
xo_list_tasks	Running XO tasks
xo_list_backups	List configured backup jobs
xo_run_backup	Run a backup job now
xo_restore_backup	Restore a VM from a backup (destructive — requires confirmed=True)
xo_rolling_pool_update	Patch every host in the pool, one at a time (destructive — requires confirmed=True)
xo_v2v_list_esxi	List reachable ESXi hosts, for a VMware migration
xo_v2v_list_vmware_vms	List the VMs on an ESXi host
xo_v2v_migrate	Migrate a VMware VM onto XCP-ng (destructive — requires confirmed=True)
xo_v2v_status	Follow a V2V migration in progress

Cross-hypervisor

Tool	Description
get_vm_usage	VM usage and cost across every declared hypervisor

Configuration (3 tools)

Tool	Description
set_cluster_rate	Set GPU cost rate for a cluster (admin only)
register_gateway	Register a new gateway (returns one-time token)
delete_gateway	Revoke a gateway

Agent Infrastructure Control Plane (12 tools)

The missing layer between your AI agents and your GPU fleet. Works with any framework (n8n, LangChain, CrewAI, Dify) — just point to the VibOps LLM Proxy.

Tool	Description
FinOps per agent
get_agent_usage	GPU cost per agent — tokens, requests, cost, GPU-hours. "Which agent costs the most?"
get_agent_usage_detail	Drill-down on one agent — daily breakdown, model distribution, cost trend
get_agent_budget	Current budget + MTD spend for an agent
set_agent_budget	Set monthly spend limit — soft alert at 80%, hard block at 100% (HTTP 429)
Model access control
get_agent_model_rules	List model access rules — which agent can use which LLM
update_agent_model_rule	Create a rule: glob patterns, deny-first. "RH agents → Mistral only"
Identity lifecycle
list_agent_identities	List machine identities for agents
create_agent_identity	Create a new machine identity (key shown once)
rotate_agent_identity	Rotate the key for an existing identity
revoke_agent_identity	Revoke an identity immediately
Dependency graph
get_agent_dependency_graph	Full org-wide graph: agent→model, agent→connector, agent→sub-agent
get_agent_dependencies	Dependencies for one agent — impact analysis before migration

Governance & Compliance (21 tools)

Tool	Description
list_anomalies	List GPU anomalies with optional cluster/status filter
get_open_anomalies	Get all currently open anomalies
resolve_anomaly	Mark an anomaly as resolved
list_compliance_controls	List compliance controls (filter by framework)
get_compliance_score	Get the compliance score for a framework (or all)
update_compliance_control	Update status, notes, or evidence URL for a control
list_compliance_reports	List generated compliance reports
generate_compliance_report	Generate a SOC 2, RGPD, or HIPAA report asynchronously
get_compliance_report	Poll/retrieve a generated compliance report
list_audit_logs	Query the immutable audit log with filters
verify_audit_chain	Verify HMAC-SHA256 integrity of the full audit chain
get_policy	Get the current organisation policy
update_policy	Replace the organisation policy (immediate effect)
list_eval_rubrics	List LLM-as-judge evaluation rubrics
evaluate_job	Trigger LLM-as-judge evaluation for a job
get_job_evaluations	Retrieve evaluation results for a job
get_ldap_config	Get LDAP / Active Directory configuration
update_ldap_config	Configure or enable/disable LDAP integration
get_siem_config	Get SIEM push export configuration
update_siem_config	Set Splunk/Datadog SIEM destination
push_to_siem	Export audit events to configured SIEM

GPU FinOps (4 tools)

Tool	Description
get_budget	Get current GPU budget and consumed spend
get_chargeback	Get chargeback breakdown by tenant for a given month
get_spend_trend	Get daily GPU spend trend (default: last 30 days)
get_waste_analysis	Identify idle GPU resources and cost optimisation opportunities

LLM Inference Proxy

VibOps includes a transparent OpenAI-compatible proxy (port 8004) that sits between your AI agents and LLM inference servers (vLLM, Ollama, TGI). Every inference request is logged with agent attribution for FinOps.

Your agents point to the proxy instead of the LLM directly:

# Before
OPENAI_BASE_URL=http://vllm:8000/v1

# After
OPENAI_BASE_URL=http://vibops-proxy:8004/v1

Add a X-VibOps-Agent-Id header to attribute costs per agent:

curl -X POST http://vibops-proxy:8004/v1/chat/completions \
  -H "X-VibOps-Agent-Id: pricing-agent-v2" \
  -H "X-VibOps-Team: supply-chain" \
  -d '{"model": "mistral:7b", "messages": [...]}'

The proxy captures: agent ID, team, model, tokens, latency, GPU cost — visible in the console FinOps dashboard and queryable via get_agent_usage.

Example prompts

"Clone my repo and deploy it on the GPU cluster."
"Deploy llama3:8b on vibops-dev with 2 replicas."
"Scale the inference deployment to 4 replicas on prod-cluster."
"What's our GPU utilisation trend over the last 7 days?"
"Show me the cost breakdown per cluster this week."
"Which clusters have open critical GPU alerts?"
"Are there any open GPU anomalies right now?"
"Scan my infrastructure and show discovered services."
"What's our AI Act compliance score and which controls are non-compliant?"
"Generate a SOC 2 report for Q1 2026."
"Verify the audit chain hasn't been tampered with."
"Which agent costs the most in GPU this month?"
"Show me the inference cost breakdown for the pricing agent."
"Which agents depend on the claude-opus-4-6 model?"
"Create a machine identity for the pricing-agent with a 1-year expiry."
"Show me the spend trend for the last 7 days and flag any waste."

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 VibOps — The AI Infrastructure Engine.