Decision Architecture

SentnelOps sits above your existing DevSecOps tools as a decision control plane. It ingests signals, reasons across domains, produces a decision, and requires human approval before any execution.

SignalsDecision AgentsDecision RecordHuman ApprovalExecution
Signals
  • CI/CD events
  • Cost anomalies
  • Security findings
  • SLO breaches
Decision Agents
  • Optimizer — cost & efficiency
  • Guardian — security & reliability
  • Arbiter — conflict resolution
Decision Record
  • Recommended action
  • Risk level
  • Expected impact
  • Full reasoning
Human Approval
  • Explicit approve / modify / reject
  • Policy-aware guardrails
Execution
  • Executed only after approval
  • Uses existing tools
  • Fully logged and reversible

SentnelOps never executes without approval

Every decision is explainable and auditable

Your tools remain the execution layer

This architecture is

  • Decision-first
  • Policy-aware
  • Human-controlled

This architecture is not

  • A replacement for your tools
  • Fully autonomous execution
  • A black-box AI system

Agents Overview

SentnelOps uses specialized AI agents, each responsible for a single dimension of DevSecOps decision-making. Together, they produce one coherent decision.

Core Decision Agents

Optimizer

Cost & Efficiency

Live
  • Evaluates resource utilization patterns
  • Identifies rightsizing and scheduling opportunities
  • Projects cost savings and efficiency impact

Does not execute changes or ignore reliability constraints.

Guardian

Security & Reliability

Live
  • Evaluates security findings and compliance requirements
  • Assesses SLO impact and blast radius
  • Produces risk assessments and go / no-go signals

Does not optimize for cost alone or allow unsafe changes.

Arbiter

Decision Resolution

Live
  • Resolves conflicts between agents
  • Applies policy priorities
  • Sequences dependent actions

The Arbiter ensures one decision, not multiple opinions.

Supporting Systems

Watcher
Early Access

Observability insights

Builder
Planned

Infrastructure execution

Conductor
Planned

Deployment orchestration

Oversight
Live

Human approvals and audit logs

Supporting systems extend signal quality and execution but do not make final decisions.

Every production decision is reviewed, approved, and logged by a human.

See a real DevSecOps decision end-to-end

Walk through how SentnelOps ingests signals, reasons across domains, and produces an explainable, human-approved decision.

No automated changes. No setup required for the demo.