OneIQ
официальныйRead-only, governed infrastructure context for any AI: VMware, Nutanix, Kubernetes and cloud, without access to production. Requires a OneIQ account.
Что можно делать с OneIQ MCP?
- Ask infrastructure questions in plain language — Query compute, virtual, storage, and container environments conversationally without writing SQL.
- Get capacity risk answers — Ask which hosts are trending toward capacity limits and receive a structured answer with drivers and recommended actions.
- Scope AI access with Agent Zones — Define named, read-only boundaries limiting which infrastructure slices an agent can query.
- Keep AI off production — Reason over infrastructure details lifted into an off-production context cloud rather than live systems.
- Receive compact, traceable answers — Get enriched, answer-shaped responses where every figure traces back to source telemetry.
Размещённый MCP-сервер
npx add-mcp 'https://mcp.oneiq.com'Устанавливается в Claude Code, Codex, Cursor и другие
Документация
Infrastructure context for every AI, agent and workflow.
OneIQ is an infrastructure intelligence platform built to put AI to work in IT. It works as your infrastructure-to-context engine, lifting your infrastructure details into a context cloud, off production, where any AI you choose can reason over them through a governed, multi-platform MCP server.
Works with your existing stack
Context Engine
AI Query
Which hosts are at risk of running out of headroom?
Answer
3 hosts in production are trending toward capacity limits within 90 days, driven primarily by memory pressure on node-7. Recommended action: rebalance workloads before the next scheduled maintenance window.
Query completed10:24 AM
Compute
4,212 hosts
Virtual
38,904 VMs
Storage
1,377 LUNs
Containers
1,860 pods
Watch how OneIQ works.
Built so the security team says yes.
Every architectural choice in OneIQ answers the questions a security team asks before an AI pilot gets anywhere near production infrastructure.
AI never touches production
OneIQ lifts infrastructure details into a governed context cloud, off production. Any AI you choose reasons over that, never over live systems, and nothing is installed on the endpoints or targets OneIQ reads.
Bounded by design
Every Agent Zone is a named, approved boundary. They are read-only, tied to a specific cluster, fleet, or infrastructure slice. An agent can't query outside its assigned zone.
Ask in plain language
Ask questions the way you'd ask a colleague, from the AI you already use. OneIQ turns each question into the right query, so nobody writes SQL and the model never holds credentials.
Answers computed, not guessed
No raw telemetry reaches the model. OneIQ resolves each query against governed data with its own enrichment logic and returns a structured answer. Every figure traces back to source telemetry.
ISO/IEC 27001 certified (BSI) Agentless - no attack surface added AES-256, zero-trust over HTTPS Per-customer encrypted boundary Secure workspaces
Production stays out of AI's reach.
OneIQ lifts your infrastructure details into a governed context cloud, so any AI you choose works from that, never from live systems. Nothing is installed on the endpoints or targets OneIQ reads, and the model never connects to them.
AI sees exactly what you configure it to see. Nothing more.
Agent zones are access boundaries you define. Each zone controls which AI agents can query which parts of your environment, what data those queries can return, and how long the zone stays active. Your team can configure zones and revoke them just as easily.
Zone definitions are fully configurable. The examples above are illustrative defaults - your team defines the exact scope of each zone during deployment.
Answers shaped for reasoning, not searching.
Without a context layer, an agent hunts through the whole environment to answer one question. OneIQ takes the question in natural language, does the finding and the computing itself, and hands the agent a compact answer. Fewer tokens spent finding it, fewer spent reasoning over it.
Point an agent at raw infrastructure
hosts: 4,212 records... vm_inventory: 38,904 rows... metrics.cpu.5m: [88,91,84,...] storage.luns: 1,377 entries... k8s.pods: 1,860 entries... events.raw: 96,500... ... everything, unscoped ...
1,000s of tokens / call
Ungoverned and over-reaching. The model burns its budget hunting the whole environment for the one thing that matters.
Point it at an Agent Zone
Zone: **"Capacity risk - Prod"** at_risk_hosts: **3** headroom: **11%** (90d ↓) top_driver: mem on node-7 recommended_action: rebalance scope: prod only · read-only
up to 98% Fewer tokens / call
Scoped, enriched, governed. Structured for reasoning, not searching.
You're in control.
Four steps, at your pace. Nobody knows your infrastructure better than the team that runs it, which makes it the lowest-risk place to run your first AI project, before extending it anywhere else.
Download
Get OneIQ Pulse context collector and install it inside your own environment.
Connect AI model
Connect the frontier or private model you already use, so it can reason over your own infrastructure.
Talk to infrastructure
Ask your first question and watch OneIQ resolve it against real, live telemetry.
Build your first Agent Zone
Define exactly what one AI agent can see: a cluster, a fleet, or a single metric.