Skip to main content

Agentic Control Plane

Most governance tools inventory your AI and draw dashboards. The Agentic Control Plane governs agents as what they actually are — actors that hold identities, spend money, call tools, and change between reviews — including the ones your teams built on other platforms.

It answers the question a security team actually asks: not what are my agents doing, but which of them can I stop, and can I prove it.

Govern agents you didn't build here​

Your teams have shipped agents on whatever platform suited them. Every one of those agents acts on your systems and spends against your budget, and until now none of them was governed by a single standard.

Point an agent's model endpoint at the Control Plane instead of its provider — a single base-URL change, no rebuild — and every call it makes inherits:

  • An assigned identity, so you know which agent acted.
  • Runtime policy on which tools it may reach, deny-by-default.
  • An enforced budget that blocks the call, rather than a report that arrives after the invoice.
  • A stop that is checked on every call — halt the agent and its next request is refused, no redeploy.
  • A signed audit trail an auditor can verify independently.

Three depths of governance​

Every agent in your register carries its depth, so the gap between seeing an agent and controlling it is never hidden.

DepthWhat it meansWhat you get
ObservedDiscovered and inventoried — including agents nobody registered by hand.Register entry, EU AI Act risk classification, audit evidence.
GatewayedRuns on its own platform, governed by yours.Everything above, plus assigned identity, runtime policy, enforced budget, a stop checked every call, full audit.
NativeBuilt on MeetLoyd.Everything above, plus cryptographic identity at creation, delegation chains, policy proven before deployment, and signed verification certificates.

Most estates start almost entirely at Observed. That's the honest starting point, and the only one from which progress is measurable — you convert the agents that matter to Gatewayed, when they matter.

Discovery​

The Control Plane can sweep your connected estate and register the agents it finds at Observed depth automatically, so the inventory stays current rather than going stale between reviews. Discovery is read-only and never changes how an agent runs — it only tells you the agent exists and puts it under governance.

Feed the control tower you already have​

If you run an enterprise AI control tower, keep it. The Control Plane exports a signed governance posture — your estate by depth, compliance classifications, verification evidence — that your tower can display and an auditor can verify was not altered in transit. Anyone can send data to a control tower; only the runtime that enforced the controls can sign for it.

Export formats include OCSF 1.3, so the feed drops into the SIEM and inventory tooling you already run.

How it's priced​

The Control Plane is priced per governed agent, not per human seat — because when agents outnumber people many times over, a seat price stops tracking the risk you're actually governing. Native agents built on MeetLoyd are covered by your platform plan and are never billed here; you pay for the foreign estate you bring under governance, at a per-agent rate that falls as the estate grows.

It's an enterprise conversation — the price depends on the size of the estate discovery finds. See Pricing & Activation and talk to us.