RadMah SentinelAvailable

Governance and audit-ready evidence for AI agents.

As teams put AI agents to work in healthcare, finance and government, RadMah Sentinel sits at runtime to govern what those agents are allowed to do — admit, route to human review, or hand off — and produces a tamper-evident, audit-ready record of every action. It answers the questions auditors ask about autonomous AI, and integrates with modern agent frameworks.

Regulated enterprises running AI agentsSecurity teamsCompliance teamsAI platform teamsAgent-framework companiesDevelopers shipping AI
RadMah Sentinel — product interface

Available today. Bring one AI-agent workflow and we'll scope a controlled evaluation with the ITLOX team.

Capabilities

What RadMah Sentinel does

The working surface — what the engine produces, controls and proves in practice.

01

Govern what agents may do, at runtime

Every action an AI agent attempts is governed against your policy before it takes effect — admitted, routed for human review, or handed to a person. Agents operate inside boundaries you set, not on trust alone.

02

Human-in-the-loop where it matters

Consequential actions pause at a review gate so a person can approve, change or stop them before they happen. No silent automation on the decisions that carry real-world consequences.

03

Tamper-evident record of every action

Each decision and action is captured in an audit-ready record designed to be tamper-evident, so what an agent did, when and on what basis stays verifiable after the fact — not scattered across tools.

04

Policy enforcement you can defend

Governance is enforced consistently at runtime rather than relying on prompts and after-the-fact logs. The result is a defensible, repeatable account of how autonomous behaviour was controlled.

05

Audit-ready evidence export

Export the reviewable evidence auditors, security and compliance teams ask for, so an oversight review or audit of your AI agents has a record to work from instead of reconstruction.

06

Integrates with modern agent frameworks

Add governance and evidence to AI agents you are already building. RadMah Sentinel works alongside modern agent frameworks rather than asking you to rebuild your stack.

07

Built for regulated oversight

Designed so regulated teams can support audit and oversight of autonomous AI — giving compliance, security and risk owners a way to see and stand behind what agents are doing.

08

One reviewable account across agents

Governance decisions, human reviews and outcomes come together in one place, so teams can oversee, audit and improve autonomous behaviour over time instead of investigating each incident from scratch.

How it works

A controlled, evidence-led flow

01

Govern

Define what your AI agents are allowed to do. At runtime, each action is admitted, routed for human review, or handed to a person, according to your policy.

02

Review

Consequential actions pause at a human-review gate. A person approves, changes or stops the action before it takes effect — keeping people in the loop where it counts.

03

Record

Every decision and action is captured in a tamper-evident, audit-ready record, so the account of what happened is preserved as it happens.

04

Prove

Export reviewable evidence for security, compliance and audit. Teams use the record to oversee, audit and improve autonomous behaviour.

Use cases

What it's used for

01

Regulated AI deployments

Put AI agents to work in healthcare, finance or government with governance and evidence in place from day one.

02

Human oversight & review

Route consequential agent actions for human review before they take effect, so people stay in control of the decisions that matter.

03

Audit & assurance

Produce the reviewable record auditors and compliance teams ask for when they question what an autonomous agent did and why.

04

Agent platform integration

Add a governance and evidence layer to an AI-agent platform or product you are already shipping.

05

Risk & compliance sign-off

Give risk owners a defensible basis to approve autonomous AI for production, rather than blocking it for lack of oversight.

06

Incident review & accountability

When something goes wrong, go to one tamper-evident record to see what the agent did, when and on what basis — instead of reconstructing it later.

Outputs & evidence

What you receive — and how it's proven

Every job ships with an evidence record, not just an output.

What you receive

  • Governance over what AI agents may do at runtime
  • Human-review gates on consequential actions
  • A tamper-evident, audit-ready record of every action
  • Exportable, reviewable evidence for security & compliance teams
  • A defensible account of how autonomous AI was controlled and overseen

Evidence & proof

  • Tamper-evident, audit-ready record of every agent action
  • Human-review gates on consequential actions
  • Reviewable evidence built for audit and oversight teams
  • Built so regulated deployments can support audit and oversight of autonomous AI
Deployment & access

Run it your way

Integrates with modern agent frameworksBuilt for regulated environmentsControlled evaluation availableAvailable today — contact ITLOX to arrange an evaluation
FAQ

Common questions

What does RadMah Sentinel actually do?

It governs what your AI agents are allowed to do at runtime — admitting an action, routing it for human review, or handing it to a person — and records every action in a tamper-evident, audit-ready form. In short, it controls autonomous AI behaviour and produces the evidence to prove how it was controlled.

Is RadMah Sentinel certified?

No — and we won't imply otherwise. It produces the tamper-evident, audit-ready evidence your own compliance and security teams need, so it strengthens your assurance work rather than standing in for a certificate.

Does it work with the agent framework we already use?

Yes — it is designed to integrate with modern agent frameworks, so you can add governance and an evidence layer to AI agents you are already building rather than replacing your stack.

How is this different from just logging to a SIEM?

Logging tells you what happened after the fact, often scattered across tools. RadMah Sentinel governs actions before they take effect — with human review where it matters — and keeps a tamper-evident record purpose-built to answer the questions auditors ask about autonomous AI.

Can a human stay in control of high-stakes actions?

Yes. Consequential actions pause at a human-review gate so a person can approve, change or stop them before they happen. There is no silent automation of the decisions that carry real-world consequences.

Is it available, and how do we evaluate it?

Yes — it is available today. Evaluations run directly with the ITLOX team: get in touch and we'll set up a controlled evaluation against one of your own agent workflows, then map the controls, review points and evidence you need.

The bigger picture

How RadMah Sentinel relates to the rest of ITLOX

RadMah AI

Governance & evidence around data generation and AI workflows.

MahCare AI

Evidence and review gates around healthcare AI workflows.

AegisWire

Policy and evidence around secure communications where supported.

Next step

Bring one AI-agent workflow you need to govern and prove. We'll map the controls, the review points and the evidence path.