How this actually works.

The Model Context Protocol is the open standard AI models use to query external systems for real information instead of guessing. Anthropic created it, but OpenAI, Google, and Microsoft all ship native support now - it's become the common language between AI agents and the tools they need to reach.

A context collector that senses telemetry across your environment and sends it to OneIQ - continuously, agentlessly, keeping a live 7-day rolling window for answering AI queries, 24×7×365. It's included at no charge with every assessment, so OneIQ always has fresh context to work from rather than a stale snapshot. Collection spans whatever you run - virtual, HCI, cloud, or bare metal - with nothing installed on target systems.

Because infrastructure telemetry wasn't built to be read by AI. It's fragmented across hybrid platforms, often on legacy gear with mixed protocols instead of clean APIs, and raw on its own. An MCP context layer can't just pass data through - it must provide context by normalizing and scoping that data into something an AI agent can query within a governed scope.

You can, and it looks fine in a demo. In production it fails because platform credentials are almost never scoped narrowly enough - the agent gets ungoverned access to everything, not just the one fact it's asking about. That turns into overreaching queries, credential sprawl, and a security team that (correctly) won't sign off.

No. A prompt instruction is a request, not an enforcement mechanism - it can be ignored, worked around, or simply fail silently. OneIQ enforces scope deterministically, in product logic. The boundary holds even if the agent misbehaves.

Raw telemetry means nothing on its own - a CPU spike could be a critical billing server or a forgotten test box, and the agent has no way to tell. Worse, it's expensive: asking through OneIQ uses up to 98% fewer tokens than querying a raw schema dump (measured July 2026).

No. It sits alongside them, reading the signals they already expose. It doesn't replace your hypervisor management, monitoring stack, or cloud consoles - it's an additional, governed layer between your infrastructure and whichever AI you choose to use.

No. OneIQ never reads the contents of your files, databases, email or applications. What it collects is context: configuration, inventory and performance telemetry, gathered read-only with nothing installed on the systems it watches. That context lives in OneIQ's governed context cloud, isolated per customer and encrypted in transit and at rest, and your AI works from it without ever touching production or its credentials. Anonymous Agent Zones can also strip identifying details like IPs and hostnames before anything reaches an external model.

A specifically governed slice of your infrastructure - a single workload, one cluster, or a migration wave. Its boundaries are hard and enforced, not by asking an LLM politely, so an agent's blast radius is confined to that zone even if it goes rogue.

No - beyond security and privacy, Agent Zones give you the benefit of isolating exactly the slice of infrastructure you want a query to run against. Examples: a capacity-planning zone covering only production clusters, a patch-risk zone scoped to just the Linux fleet, a cost zone covering only cloud + on-prem workloads flagged for right-sizing, or a migration-wave zone limited to the hardware in that specific wave. Each zone becomes a focused lens - the agent reasons only over the slice relevant to the question, not everything else.

Very little. An agent can only see its own zone, and that zone holds a copy of your infrastructure's context in OneIQ's cloud, not production itself. It has no credentials and no route back, so it can't reach your real systems. And every request it makes is logged.

The part of OneIQ that does the heavy lifting itself. When an agent asks a natural-language question, the inline retrieval layer resolves it against your already modeled and enriched environment and hands back compact, structured facts - instead of making the AI burn tokens parsing raw data or API documentation to find the answer itself.

Because raw telemetry on its own is just a list of what you own - it can't tell you what's three generations out of date, what your power bill implies, or what a cloud alternative would cost. OneIQ joins that telemetry with real-world context so the answer is decision-ready, not another spreadsheet to interpret.

They add the facts your infrastructure can't supply on its own: current cloud pricing, exchange rates, and hardware and software lifecycles. OneIQ joins them to your context automatically, so your AI can answer "how old is this hardware?" or "which hosts are running software that's out of support?" without searching the web.

Once OneIQ is approved in your environment, your first proof of concept, Ask Your Infrastructure, is ready in under 30 minutes. From there you have the OneIQ portal for a single view across your infrastructure, Agent Zones to set exactly what each AI or team can see, and an MCP-ready connection so the AI you already use can put that context to work.

Collection can begin immediately, normalization and enrichment run within minutes, and scoping your first zone in the UI typically means a live, governed Agent Zone in under 30 minutes - no advanced AI engineering team or capital hardware budget required.

Yes - that's the "my AI talks to your AI" model. As vendors and partners expose their own governed AI interfaces, an external AI can reason over exactly the source-masked context you choose to share, and nothing more, to do things like coordinate a migration or answer an RFP - without either side exposing raw infrastructure.

Without a governed layer in between, yes - that would be a real risk. With OneIQ, the external AI only ever sees the specific source-masked Agent Zone you've defined, enforced the same way internal access is enforced. The rails are already laid; you're not opening a new door for every partner.