AegisQ AI Sentinel
Early access · v3.9.0 · Software + Physical AI

AegisQ AI Sentinel

Govern every model and agent.

Universal behavioral monitoring for software AI agents and physical AI systems. Behavioral charters, kinematic safety charters, real-time enforcement (L1–L5), sensor attestation, fleet coordination, and quantum-resistant compliance proof. One platform for both surfaces.

LLM agents LangChain · CrewAI OpenAI · Anthropic Collaborative arms Industrial arms AMRs Humanoids Surgical robots Drones & UAVs
The problem

Autonomous AI is shipping faster than the governance for it.

Software agents on LangChain, CrewAI, OpenAI, and Anthropic make autonomous decisions without guardrails or audit trails. Collaborative robots, AMRs, and drones operate in shared human spaces with kinematic safety risks software alone can't govern. When connectivity drops, cloud-based governance disappears. AI Sentinel is one platform for both surfaces — software and physical AI — under one compliance framework, with cryptographic, quantum-resistant proof.

Software AI governance

Runtime monitoring for LLMs, agents, and ML pipelines.

Six core capabilities define what each agent is allowed to do, detect when it drifts, and respond proportionally — every action logged and signed.

Software AI

Behavioral charter engine

Define what each agent is allowed to do, spend, and access. Charters are the source of truth — every detection and response references them.

Software AI

Anomaly detection suite

Behavioral profiling, capability drift, self-modification monitoring, intent inference, and digital-twin comparison — multi-layered, in real time.

Both

Graduated response (L1–L5)

From passive observation to emergency halt. Five severity levels with full audit trails — proportional response for software agents and robots alike.

Software AI

Anti-circumvention

Liveness probes, canary rules, and alert-flood detection ensure monitored systems can't tamper with or evade Sentinel's monitoring boundary.

Both

Cryptographic attestation

ML-DSA-65 (FIPS 204) signed compliance reports — tamper-evident, quantum-resistant evidence that every system operated within charter.

Software AI

Agent trust & identity

DID-based identity, W3C Verifiable Credentials, and trust scoring with reputation decay — a verifiable reputation system per agent.

Also included

FinOps engine

Cost attribution and budget enforcement per agent, granular spending controls, financial guardrails.

Decision provenance

Full delegation graphs, authorization chains, and regulatory export for audit compliance.

Certification suite

Automated test suites, risk scoring, and actuarial data for AI-insurance readiness.

Agent lineage

Track agent spawning, permission inheritance, and parent-child relationships across runtimes.

MCP & supply-chain governance

Govern the agent supply chain, not just the agent.

As of v3.9.0, AI Sentinel governs the Model Context Protocol surface that modern agents depend on — the tools they call and the telemetry they emit — with all phases of the OWASP Data Security for GenAI (DSGAI) 2026 framework implemented.

Software AI

MCP event governance

A permission-scoped MCP event sink governs which agents and tools may report telemetry. Charter violations resolve into typed INTERVENE / HALT action records that emitters retrieve and enforce per tenant.

Software AI

CVE-aware MCP rules

MCP detection rules are backed by live OSV.dev advisory data, flagging known-vulnerable MCP server and tool versions across the environments AI Sentinel governs.

Both

OWASP DSGAI 2026 coverage

All phases of the OWASP Data Security for GenAI framework are implemented — runtime content inspection and GenAI data-loss controls across the telemetry pipeline.

Physical AI governance

Kinematic safety, sensor attestation, fleet coordination.

Six robot families. Six governance capabilities. One control plane. Built for environments where safety boundaries are physical, not just behavioral.

Collaborative arms

UR, Franka, KUKA iiwa cobots operating alongside humans.

Industrial arms

KUKA KR, ABB IRB, Fanuc — high-payload production robots.

Mobile robots (AMRs)

Autonomous mobile robots for warehouse, logistics, and facility navigation.

Humanoids

Bipedal humanoids in manufacturing, service, and research.

Surgical robots

Precision medical and surgical systems in clinical settings.

Drones & UAVs

Aerial drones for inspection, delivery, surveillance, and mapping.

Physical AI

Kinematic safety charters

Joint limits, velocity caps, force thresholds, proximity zones, and geofence boundaries — enforced in real time with pre-execution validation.

Physical AI

Sensor attestation

Merkle-proof integrity, drift detection, and tamper alerts for every sensor on every robot. Cryptographic proof of sensor health at measurement time.

Physical AI

Physical digital twin

URDF-based robot models for pre-execution validation, collision prediction, and what-if simulation before motion commands hit hardware.

Physical AI

Fleet coordination

Shared workspace monitoring, emergent risk detection across robot types, fleet-wide emergency stop. Govern multi-robot environments as a unified system.

Physical AI

Cryptographic lease system

Robots continue safe operation during connectivity loss with time-limited cryptographic leases. When leases expire, hardware failsafes engage automatically.

Physical AI

Safety event pipeline

Real-time safety event logging, incident classification, and automatic regulatory reporting. 7-year retention for OSHA and ISO 10218 compliance.

Three operating modes

Governance never stops. Even when the network does.

Connected

Full cloud governance

Cloud-based evaluation, real-time charter updates, fleet coordination, and streaming telemetry. Optimal governance with complete visibility.

Degraded

On-robot enforcement

On-robot agent with cached charters and store-and-forward telemetry. Safety enforcement continues locally when connectivity is limited or intermittent.

Disconnected

Cryptographic leases

Cryptographic lease system with hardware failsafe integration. Local enforcement, zero network dependency, leases expire to fail-safe state.

Ethernet WiFi 5G / 4G LTE Satellite (LEO/GEO) Mesh radio Tethered Air-gapped
Graduated response system

Five levels. Proportional. Auditable.

Every response action is logged with full audit trail. The same five-level scale applies to software agents and physical robots.

L1Observe

Observe

Continuous observation with zero intervention. All activity and telemetry logged but no restrictions applied. Baseline profiling for agents and robots.

Software AI + Physical AI
L2Alert

Alert

Real-time notifications on anomalies and charter violations. Teams receive alerts for review and manual intervention without system lockdown.

Software AI + Physical AI
L3Throttle

Throttle

Reduce agent rate limits or restrict robot velocity / force envelopes. Systems remain operational within tighter safety boundaries.

Software AI: rate limits · Physical AI: velocity / force caps
L4Intervene

Intervene

Pause agent execution or command robot to safe-hold position. Systems suspended pending human review and remediation.

Software AI: suspend execution · Physical AI: safe-hold position
L5Halt

Halt

Emergency shutdown of the agent and all child processes, or fleet-wide emergency stop for robots. Reserved for critical safety breaches.

Software AI: terminate + children · Physical AI: fleet e-stop
Supported platforms

Software frameworks and robot platforms — one governance layer.

OpenAI (GPT, Assistants) Anthropic (Claude) LangChain CrewAI Generic HTTP UR (Universal Robots) KUKA (iiwa / KR) ABB (IRB Series) Fanuc ROS 2 / DDS
Compliance & security

Quantum-resistant. Multi-tenant isolated. Patent-protected.

AegisQ AI Sentinel is engineered design-aligned with the regulatory frameworks below. SOC 2 Type II audit is targeted for H1 2027 and ISO 27001 certification will follow. Until those certifications are complete, framework alignment is by design and self-attestation only.

Quantum-resistant encryption

ML-DSA-65 (FIPS 204) JWT signing, ML-KEM hybrid TLS for pod-to-pod communication. Future-proof against CRQCs.

5-tier tenant isolation

From shared row-level security to fully isolated clusters. Choose the level that matches your security and regulatory obligations.

14 provisional patents

Protected innovations across behavioral governance, kinematic safety, quantum-resistant attestation, and multi-robot coordination.

SOC 2 Type II GDPR CCPA HIPAA ISO 27001 ISO 10218 IEC 62443 EU AI Act NIST AI RMF OWASP DSGAI 2026
10
Software AI capabilities
6
Robot families
L1–L5
Graduated response
3
Operating modes
14
Provisional patents
Pricing

Priced per agent. Scale from a laptop to a fleet.

A base platform fee with included agents, then volume-discounted agent blocks as you scale. Physical AI monitoring is an add-on from the Professional tier up.

Free (Dev)

$0

2 agents · 5,000 events/day · 1 tenant · Tier 1 shared isolation · 7-day audit retention · community support. For evaluation and development.

Get started

Starter

$149 / mo

10 agents · 100K events/day · Email + Slack alerts · Tier 1 isolation · 30-day retention · email support (48hr). For startups and small teams.

Start a trial

Business

$1,799 / mo

200 agents · 5M events/day · Tier 3 schema isolation · SAML SSO · behavioral profiling · FinOps + certification API · 365-day retention.

Talk to sales

Enterprise

From $5,000 / mo

500+ agents · unlimited events · Tier 4 dedicated database · custom detection rules · 99.9% SLA · 1–7yr retention · 24/7 dedicated support.

Talk to sales

Enterprise Gov

From $10,000 / mo

Everything in Enterprise · Tier 5 cluster isolation · CMEK · FedRAMP High / ITAR / SOC 2 Type II · WORM audit · self-hosted or dedicated GCP project.

Talk to sales

Additional agents from $0.50–$2.50/agent in volume-discounted blocks. Annual plans save ~17% (2 months free).
Physical AI add-ons (Professional+): Collaborative arm $299/mo · Industrial arm $499/mo · AMR $199/mo · Humanoid $999/mo · Surgical $1,499/mo · Drone $149/mo

Govern every model and agent.

30 minutes to walk through behavioral charters, kinematic safety, and L1–L5 response on your own agents and robots.