TroutTrout

Agentless Zero Trust for OT Zero Trust Without Agents on the Asset.

Most OT devices cannot run security agents. PLCs, HMIs on Windows XP, ten-year-old engineering workstations, and SCADA servers either reject endpoint software or run it at the cost of vendor support. The April 2026 CISA guidance endorses agentless deployment for legacy OT and recommends pairing it with network-layer enforcement. This page covers why, what, and how.

Why Agents Fail in OT.

IT Zero Trust assumes the device can host an agent that verifies posture, enforces policy, and reports telemetry. OT cannot make that assumption. Proprietary real-time operating systems do not allow third-party software. Deterministic control loops cannot tolerate the timing jitter an agent introduces. Vendor warranties explicitly void if endpoint software is installed. The agentless question is not whether to deploy without agents. It is whether the network-layer enforcement that replaces them is strong enough.

What CISA Says

The April 2026 Position.

CISA, the Department of War, DOE, FBI, and Department of State jointly published Adapting Zero Trust Principles to Operational Technology on April 29, 2026. On the agent question (p.17), the guidance is direct: agents require extensive compatibility testing and can impact system warranties, while passive monitoring alone misses abuses in remote-access sessions. The endorsed posture is agentless deployment combined with active enforcement at the network boundary. This document is the highest-authority federal guidance to date on agentless Zero Trust for OT.

Read our full analysis of the CISA guidance
Compensating Controls Catalog

What Replaces the Agent.

When the asset cannot host security software, the network must provide what the agent would have provided. Each capability below is what CISA calls a compensating control, mapped to specific Access Gate functionality.

Identity-Bound Access
Agent equivalent

Endpoint MFA prompt

How Access Gate Provides It

MFA enforced at the network boundary against your IdP. The asset never sees the user identity directly. Each session is bound to a named user, not a service account or shared credential.

Session Authorization
Agent equivalent

Local policy enforcement

How Access Gate Provides It

Per-asset, per-protocol, per-session policy. A single user can be authorized to read a Modbus register on PLC A and denied write access to PLC B in the same authentication context.

Audit and Session Recording
Agent equivalent

Endpoint event log

How Access Gate Provides It

Every session is logged with user identity, timestamp, asset, protocol, and full payload when configured. Logs are tamper-evident through hash chaining and FIPS-validated signing. 90-day retention default.

Microsegmentation
Agent equivalent

Host-based firewall rules

How Access Gate Provides It

Overlay-based microsegmentation enforces per-asset boundaries on top of the existing physical network. The Layer 2 topology does not change. Safety PLCs and control PLCs can sit on the same VLAN with isolated reachability.

Encryption for Legacy Protocols
Agent equivalent

Endpoint TLS stack

How Access Gate Provides It

Modbus, DNP3, OPC UA, EtherNet/IP, and other industrial protocols are wrapped in FIPS-validated TLS at the proxy boundary and forwarded natively to the asset. The asset does not need to support TLS.

Anomaly Detection on Session
Agent equivalent

Endpoint EDR

How Access Gate Provides It

Session-level anomaly detection on user, time, command sequence, and payload patterns. Flagged sessions can be blocked at the appliance, not just alerted downstream.

Just-In-Time Access
Agent equivalent

Local elevation prompt

How Access Gate Provides It

Vendor and contractor sessions are scoped, time-bound, and revocable. Access closes automatically when the session ends. No persistent VPN tunnels, no standing credentials.

FAQ

Frequently Asked Questions

What does agentless Zero Trust mean for OT?
Agentless Zero Trust enforces identity-based access, segmentation, and audit logging without installing software on the protected asset. Enforcement happens at a network boundary in front of the asset, not inside the asset. This is the posture CISA endorses for legacy OT, where most devices cannot host agents.
Why can't OT devices run security agents?
Three reasons. First, proprietary real-time operating systems on PLCs and industrial controllers do not support third-party software. Second, deterministic control loops cannot tolerate the timing jitter an agent introduces. Third, OT vendor warranties typically void if endpoint software is installed.
Does CISA require agentless Zero Trust for OT?
The April 2026 CISA guidance is recommendation, not regulation. It does not impose a compliance deadline. It is the highest-authority federal publication to date on Zero Trust for OT and informs the controls auditors and assessors will look for under CMMC, NIS2, and DoD frameworks that do have compliance force.
Is passive monitoring enough?
Per CISA, no. Passive monitoring catches anomalies after the fact but does not block malicious commands in active remote-access sessions until they are sent. The endorsed posture combines agentless deployment with active enforcement at the network boundary that can deny a session in real time, not just log it.
How is agentless Zero Trust different from a firewall?
A firewall asks: is this port allowed between these two subnets? Agentless Zero Trust asks: is this specific user, on this specific session, allowed to send this specific command to this specific asset, right now? The firewall enforces at Layer 4. Zero Trust enforces at the application layer, with identity binding, session attribution, and full payload logging.
How does Access Gate enforce Zero Trust without an agent?
Access Gate is an on-premise appliance or VM that proxies access to OT assets at a network boundary. It runs adjacent to the network, not inline. Every session terminates at the appliance, where authentication, authorization, and full session recording happen before traffic reaches the asset. The asset itself never installs software, never sees the user identity directly, and never knows the access control exists.
Next Step

See Agentless Zero Trust in Action.

Request a 30-minute working session. We will walk through your environment, identify which assets cannot host agents, and show how Access Gate enforces Zero Trust at the network boundary in front of them.

Request a Working Session

Related