AX Assurance.

A control layer around your agents: audit trail, policy gate, escalation — governance that makes Agentic AI enterprise-ready.

DACH-wide

Book an initial call

An agent that writes or executes needs a control authority — otherwise "autonomous" is just another word for "unsupervised". AX Assurance wraps around your agents as a control layer: a policy gate checks every call beforehand, an audit trail records it afterwards, and critical findings are escalated for human review.

ax-assurance — Gate & Audit Trail

# Policy gate — before every agent call
  ✓ ALLOWED    ax_test_pipeline   sqlite:///app.db
  ✗ BLOCKED  ax_test_pipeline   db_url points to a protected environment ('prod')
  ✓ ALLOWED    ax_review_repo     /srv/app

# Audit trail — after every call (input as hash only)
  2026-06-13T07:18Z  ax_test_pipeline  ok=true   verdict=approve
  2026-06-13T07:18Z  ax_test_pipeline  ok=false  policy: protected environment
  2026-06-13T07:18Z  ax_test_pipeline  ESCALATION: pytest failed → human review

The gate above is real: it permits a local database run but blocks a production database, writes an audit entry (inputs as a hash only, not in plain text) and escalates a failing test run. To be honest about scope: for full coverage of the OWASP Agentic Top 10 we build on established open source — the Microsoft Agent Governance Toolkit — rather than reinventing a governance engine. Our IP is the curated configuration: which guardrails for which agent.

Policy gate (before the call)

Every agent call is checked against rules — e.g. no production DB, only permitted paths. Violations are blocked, not executed.

Audit trail (after the call)

Complete logging of every call. Inputs are stored as a hash only — evidence without secrets leaking.

Escalation

Critical findings — failing tests, critical review findings — are flagged and escalated for human approval instead of running through.

OSS foundation

For full OWASP Agentic Top 10 coverage, built on the Microsoft Agent Governance Toolkit (open source) — no in-house engine.

Policy gate before every agent call — configurable allowlists and prohibitions
Tamper-resistant audit trail: every call logged, inputs as a hash only
Automatic escalation of critical findings for human review
Applies across all AX agents (Code Review, Test Agent, Optimize)
Honest OSS base: full coverage via the Microsoft Agent Governance Toolkit — our IP is the configuration

Does this replace a SIEM or our existing compliance?

No. AX Assurance sits specifically around your AI agents — policy gate, audit, escalation at the agent level. You can feed the audit trail into your existing SIEM or compliance reporting; it does not aim to replace it.

Does the audit trail store our code or our inputs?

No. Inputs are logged as a hash plus size only, not in plain text. You have the evidence that and when a call ran — without secrets ending up in the log.

What does it cover today, and what comes via open source?

Today the gate, the audit trail and the escalation run in production. For full OWASP Agentic Top 10 coverage we integrate the Microsoft Agent Governance Toolkit — established, open source, not reinvented by us.

How does it hook into our MCP agents?

The gate runs before, the audit after every tool call the agent makes — transparent to the agent itself. A blocked call is returned to the agent as a controlled error, not silently executed.

Related roles & solutions

Goes together

Product AX Code-Review-Agent The agents that run beneath the control layer. Product AX Optimize Cost governance as a complement to security governance. Overview Agent-Lösungen The full AX Suite and tailored solutions.

AX Assurance — does this fit your situation?

In a free initial call we clarify whether this service addresses your problem — or whether another route gets there faster. Honest assessment included.

Book an initial call