Customer-resident deployment
AgentGuardian runs in the customer's AWS environment, reducing vendor-risk concerns and supporting regulated deployment models.
Glacien · AgentGuardian — Agentic AI, governed.
AgentGuardian
Govern every AI agent in your AWS estate.
AgentGuardian is Glacien’s AWS-native AI agent governance, security, operations, and evidence platform for regulated enterprises. Deployed inside the customer’s AWS environment, it helps organisations discover every AI agent, classify risk, enforce policies, red-team high-risk agents, monitor production behaviour, control cost, and generate audit-ready Evidence Packs for risk, compliance, internal audit, and supervisory review.
Discover · Govern · AssessOperate · Monitor · EvidenceCustomer-resident · AWS-native
Defensible to every regulator and audit body that reviews your AI agents.
MAS, APRA, and internal-assurance coverage out of the box — your CRO no longer has to translate between regulator language and what your agents actually do.
For each jurisdiction or assurance function, AgentGuardian names the standard, what you can prove, and the artefacts that prove it.
Singapore — MAS
Prove your AI agents meet MAS expectations on risk management and the FEAT principles.
Covers
- MAS AI risk management expectations
- MAS FEAT principles (Fairness · Ethics · Accountability · Transparency)
Australia — APRA
Demonstrate AI risk control under APRA’s AI observations and operational & information-security standards.
Covers
- APRA AI risk observations
- APRA CPS 230 (operational risk)
- APRA CPS 234 (information security)
Internal assurance
Give internal audit, technology risk, and model risk teams the audit-ready evidence they need on demand.
Covers
- Internal audit reviews
- Technology risk management
- Model risk governance
The product
One platform. Six capabilities.
Each capability is a discrete module inside a single platform, sharing identity, traces, and evidence.
Discover
Capability 01
Continuously discover AI agents, MCP servers, tools, model endpoints, service identities, data access, owners, and business use cases across the customer’s AWS estate.
Outputs include
- AI Agent Inventory
- AI Bill of Materials
- Shadow-agent report
- Ownership map
- Tool and data access map
- Risk-tiering view
Govern
Capability 02
Apply enterprise policies for data access, tool usage, human approval, prohibited actions, exception handling, and risk ownership across AWS-hosted agent workflows. AgentGuardian moves AI controls out of policy documents and into enforceable governance workflows.
Outputs include
- Policy library
- Control map
- Exception register
- Approval workflow
- Ownership model
- Policy enforcement record
Assess
Capability 03
Red-team agents before production and on a recurring basis. AgentGuardian tests against prompt injection, jailbreaks, tool misuse, data leakage, retrieval poisoning, identity and permission bypass, human-approval bypass, and guardrail weakness.
Outputs include
- Findings report
- Risk scorecard
- Security test results
- Mitigation roadmap
- Re-test evidence
- Production-readiness recommendation
Operate
Capability 04
Run AI agents like production systems, not demos that survived release. AgentGuardian tracks performance, latency, quality, token usage, cost attribution, tool execution, failure patterns, release history, and improvement actions.
Outputs include
- Operations dashboard
- Cost and usage report
- Quality scorecard
- Release log
- Failure analysis
- Monthly improvement backlog
Monitor
Capability 05
Detect runtime drift, hallucination risk, abnormal tool usage, identity anomalies, prompt-injection signals, memory poisoning, policy violations, and autonomous-action cascades. AgentGuardian turns agent behaviour into observable, traceable, reviewable operational evidence.
Outputs include
- Runtime alerts
- Incident cases
- Trace evidence
- Anomaly reports
- SOC integration
- Escalation history
Evidence
Capability 06
Generate signed Evidence Packs mapped to internal controls, MAS/APRA expectations, audit requirements, and AI governance frameworks. The audit and regulatory assurance layer — showing what agents exist, who owns them, what they access, how they were tested, and what evidence supports the control posture.
Outputs include
- Signed Evidence Pack
- Control-coverage map
- Audit-ready record
- Exception history
- Quarterly assurance report
- AI Risk Committee report
What you get.
Six executive-grade artefacts. Each engagement produces working software in your AWS account, plus these named deliverables in your CRO’s and CIO’s hands.
01
Enterprise AI Agent Inventory + AI Bill of Materials
A complete, continuously-refreshed inventory of every AI agent in your AWS estate — owners, tools, data access, model endpoints, identities — published as your AI Bill of Materials.
02
Risk-tiering & shadow-agent report
Every agent classified by materiality (Impact · Complexity · Reliance). Previously-unknown shadow agents named, owned, and brought under governance.
03
Red-team findings & mitigation plan
Prompt-injection, jailbreak, tool-misuse, data-leakage and retrieval-poisoning test results on your top-tier agents — with mitigation roadmap and re-test evidence.
04
Production-readiness scorecard
Operational risk assessment of agents already in production: observability gaps, cost, quality, failure patterns, named owners, and an activation roadmap.
05
Signed Evidence Pack — quarterly
Examiner-ready pack mapped to MAS, APRA and your internal control framework. Counter-signed, hash-chained, time-stamped. The artefact your CRO hands to a regulator.
06
90-day governance roadmap
Phased remediation plan with named owners, dated deliverables, and quarterly executive checkpoints. The plan your AI Risk Committee tracks against.
Quick start
Start with an assessment.
Three fixed-scope engagements. Pick the one that fits where your AI agent estate is today.
Path 01 · Sprawl
Agent Sprawl Assessment
Includes
- Agent discovery scan
- AI Agent Inventory
- Shadow-agent report
- AI Bill of Materials
- Risk tiering & ownership map
- Materiality assessment
- 90-day governance roadmap
Path 02 · Security
Agent Security Assessment
Includes
- Prompt injection testing
- Jailbreak testing
- Tool misuse testing
- Data leakage testing
- Retrieval poisoning testing
- Identity & permission review
- Human-approval bypass testing
- Mitigation & re-test plan
Path 03 · Production
Production Readiness Assessment
Includes
- Runtime observability review
- Cost & token usage baseline
- Quality & evaluation review
- Tool execution review
- Failure & exception analysis
- Operational risk scorecard
- AgentOps activation roadmap
From assessment to platform.
After the starting engagement, activate AgentGuardian inside your AWS environment.
Four expansion shapes — pick the one that matches the next business outcome.
Expansion 01
Single capability
Start with one priority area — Discover, Govern, Assess, Operate, Monitor, or Evidence.
Expansion 02
Multi-capability bundle
Combine capabilities around a specific business need — regulatory readiness, production assurance, or agent security.
Expansion 03
Enterprise platform
Govern AI agents across business units, production workflows, use cases, and risk tiers.
Expansion 04
Quarterly AI Assurance
Refresh red-team testing, control evidence, governance reporting, and production assurance on a recurring cadence.
How AgentGuardian supports audit & regulatory expectations.
From requirement to capability to evidence.
For every audit or regulatory need, AgentGuardian names the capability that addresses it and the artefact it produces.
Regulatory & audit need
AgentGuardian capability
Evidence produced
AI agent inventory and ownership
Discover
AI Agent Inventory, AI Bill of Materials, owner register
AI risk classification and materiality
Discover + Govern
Risk-tiering report, materiality assessment, high-risk agent register
Board and senior management oversight
Govern + Evidence
Governance dashboard, approval history, AI Risk Committee report
Policy and control enforcement
Govern
Policy map, control library, exception register, approval rules
Non-human identity governance
Govern + Monitor
Agent identity map, permission scope, access trail, anomaly report
AI lifecycle controls
Govern + Assess + Operate
Pre-production assessment, release readiness, evaluation results, change history
Security testing and red-teaming
Assess / AgentShield
Prompt-injection tests, jailbreak tests, data-leakage tests, remediation record
Ongoing monitoring and drift detection
Monitor
Runtime alerts, drift reports, hallucination signals, anomaly cases
Operational resilience and incident response
Operate + Monitor
Incident runbook, failure analysis, escalation history, recovery evidence
Cost and usage accountability
Operate / AgentOps
Token usage, model cost, agent cost attribution, FinOps report
Internal audit readiness
Evidence
Signed Evidence Pack, control map, audit trail, exception history
Regulatory review readiness
Evidence
MAS/APRA-mapped Evidence Pack, quarterly assurance report
Runs inside your AWS environment.
Deployed as an AWS-native SaaS solution — customer data, keys, and logs remain under customer control.
Identity, runtime visibility, monitoring, policy enforcement, audit, encryption, and evidence storage all run on AWS-native services.
AWS capability
Role in AgentGuardian
Amazon Bedrock AgentCore Runtime
Production-grade agent execution foundation
Amazon Bedrock AgentCore Identity
Agent identity, permissions, and on-behalf-of access
Amazon Bedrock AgentCore Gateway
Tool and API access control layer
Amazon Bedrock AgentCore Observability
Traces, debugging, monitoring, and runtime visibility
Amazon Bedrock AgentCore Evaluations
Continuous quality, safety, and performance assessment
Amazon Bedrock Guardrails
Content safety, sensitive data controls, and grounding checks
AWS CloudTrail
Audit trail and control history
Amazon CloudWatch
Operational metrics, logs, traces, and alerts
AWS KMS
Customer-managed encryption
Amazon S3 Object Lock
Tamper-resistant evidence storage
AWS Security Hub
Security findings and SOC integration
AWS Lake Formation + Athena
Evidence lake and governed reporting
Why customers choose AgentGuardian.
Five reasons regulated enterprises pick AgentGuardian.
01
02
Built for regulated AI
Designed for organisations where AI agents must be secure, auditable, controlled, and defensible to risk, compliance, internal audit, and supervisory stakeholders.
03
One platform, six capabilities
Combines agent discovery, policy enforcement, red-teaming, production monitoring, cost control, operational assurance, and evidence generation.
04
Delivered, not just advised
Glacien deploys the platform, configures the controls, integrates the tooling, and supports the operating cadence with the customer.
05
AWS-native foundation
Designed around AWS-native services, making it easier for AWS-based enterprises to adopt, operate, and govern agentic AI.
About Glacien
We help regulated enterprises build, govern, and operate agentic AI on AWS.
We work with organisations where AI agents must be secure, auditable, production-ready, and defensible to risk, compliance, and supervisory stakeholders.
Singapore-headquartered. AWS Select Partner with Agentic AI specialisation. Built for enterprise agentic AI.
Agentic AI, governed.
Ready to govern every AI agent in your AWS estate?
Book a 30-minute walkthrough. We will show the agent inventory model, sample Evidence Pack, AWS deployment pattern, and 90-day governance roadmap. No slides that waste your time.