The First Governed
Autonomous Employee
Hermes is not a chatbot. Hermes is the first governed autonomous employee running on AIF infrastructure — demonstrating every capability AIF provides to customers.
What Makes UFA Different
Every AI system claims autonomy. UFA is the only one that proves it — with constitutional governance, sealed evidence packages, and independent replayability.
Constitutional Runtime (Frozen v3.1)
Four-gate review at every lifecycle stage: Architecture (20%), Operational (25%), Trust (25%), Evolution (15%), Utilization (15%). No governance drift possible.
Temporal as Default Substrate
All durable work executes on Temporal. Automatic retries, compensation, scheduling, signals, and deterministic replay. Anti-pattern detection prevents cron wrappers and glorified job queues.
Sealed Evidence Packages
Every execution produces an immutable bundle: Receipt + Replay ID + Workflow History + Activity Manifest + OTel Trace + Metrics + Policy Snapshot + Identity + Search Attributes + Replay Result + Constitution Verdict.
Principal Attribution on Every Start
Constitutional interceptors auto-collect identity, policy, mission, workflow version, worker version, namespace, search attributes, and Temporal Build ID. No anonymous execution.
Operational Status
Mission Registry — Autonomous Operating Department
Every recurring operational function of AIF is a governed mission. Constitution review → Temporal execution → Evidence package → Evaluation → Learning loop.
MissionExecution
Core workflow orchestration with constitutional gates
AgentLifecycle
Agent spawn, health, upgrade, decommission
ReceiptVerification
Automated receipt validation and replay
CustomerAcquisition
Lead → proposal → contract → onboarding
ContentFactory
Technical content production pipeline
AutonomousCompany
Back-office: billing, compliance, reporting
ResearchIntelligence
Competitive monitoring, thesis validation
SelfImprovement
Capability routing, policy adjustment, automation
CapitalAllocation
Budget optimization, ROI tracking
Federation
Multi-tenant coordination, cross-agent workflows
TemporalOptimization
Daily 05:00 UTC · utilization engine + primitive optimizer
UFA Architecture Overview
Three-layer architecture enforcing AIF Trust Invariants at every boundary.
┌─────────────────────────────────────────────────────────────────────────────┐
│ INTENT LAYER │
│ ┌──────────────┐ ┌──────────────────┐ ┌──────────────────────────────┐ │
│ │ Intent │ │ Mission │ │ Human/Policy │ │
│ │ Compiler │──▶│ Proposal │──▶│ Approval Gates │ │
│ │ (NL → Spec) │ │ Constitutional │ │ (4-Gate Review) │ │
│ └──────────────┘ │ Review │ └──────────────────────────────┘ │
│ └──────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ EXECUTION LAYER (TEMPORAL) │
│ ┌──────────────────────────────────────────────────────────────────────┐ │
│ │ Constitutional Interceptors │ │
│ │ Auto-collects: Identity │ Policy │ Mission │ Version │ Namespace │ │
│ │ Search Attributes │ Build ID │ Principal Attribution │ │
│ └──────────────────────────────────────────────────────────────────────┘ │
│ │ │
│ ┌─────────────────┐ ┌─────────────────┐ ┌────────────────────────────┐ │
│ │ P0 Workflows │ │ P1 Workflows │ │ P2 Workflows │ │
│ │ MissionExec │ │ CustomerAcq │ │ SelfImprovement │ │
│ │ AgentLifecycle │ │ ContentFactory │ │ CapitalAllocation │ │
│ │ ReceiptVerify │ │ AutonomousCo │ │ Federation │ │
│ │ │ │ ResearchIntel │ │ TemporalOptimization │ │
│ └─────────────────┘ └─────────────────┘ └────────────────────────────┘ │
│ │ │
│ ┌──────────────────────────────────────────────────────────────────────┐ │
│ │ Activity Heartbeats │ Compensation (Saga) │ Retry Policies │ │
│ │ Continue-As-New (long-running) │ Signals (approval) │ Schedules │ │
│ └──────────────────────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────────────┐
│ TRUST LAYER │
│ ┌──────────────────┐ ┌──────────────────┐ ┌──────────────────────────┐ │
│ │ Evidence │ │ Receipt │ │ Observatory │ │
│ │ Packages │ │ Verification │ │ (13-Dimension Health) │ │
│ │ (Immutable, │ │ (Deterministic │ │ • Availability │ │
│ │ Sealed) │ │ Replay) │ │ • Latency │ │
│ └──────────────────┘ └──────────────────┘ │ • Throughput │ │
│ │ • Error Rate │ │
│ ┌──────────────────┐ ┌──────────────────┐ │ • Durability │ │
│ │ PDP Authority │ │ Audit Engine │ │ • Replayability │ │
│ │ (OPA/Rego) │ │ (Autonomous) │ │ • Constitutional │ │
│ └──────────────────┘ └──────────────────┘ │ Compliance │ │
│ │ • Identity Integrity │ │
│ │ • Evidence Completeness │ │
│ └──────────────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘Every Claim Above Is
Replayable & Auditable
Independent auditors can verify any UFA execution by replaying the sealed evidence package. No trust required — only verification.