Execution governance for AI agents · ProvenanceCode Agent OS

The control plane between AI reasoning and real-world execution.

The missing layer between AI intent and execution. Agent OS is a governance and audit runtime that sits between your AI agent’s reasoning and real-world action. It intercepts every agent decision, evaluates it against policy, creates a provenance record, and either executes, gates, or blocks — before anything touches your infrastructure.

AWS Marketplace
ISO 42001 policy
Amazon Bedrock
2-input setup
Agent OS · policy engine
agent.execute("deploy api-service to prod")
Agent OS: planning actions...
Generating provenance record...
Risk level: HIGH · production deploy
Policy: requires_human_approval = true
Awaiting approval... // SMS sent
✓ Approved by: k.murphy@company.com
✓ Executing via aws.ecs executor
✓ Action complete. Signed provenance stored.
100%
Actions auditable
Risk tiers
<50ms
Policy eval overhead
Executor types
🔍
Audit

Every action logged with full provenance — who, what, why, when.

⚖️
Policy

ISO 42001-based rules gate actions before execution — deterministically.

🔔
Control

Human-in-the-loop approval for high-risk operations. The agent waits.

☁️
AWS Marketplace

Subscribe, enter your Bedrock or OpenAI key plus ProvenanceCode API key — default policy applied, containers online.

📐
Risk-weighted actions

Low ×1 · Medium ×3 · High ×10 — fair metering aligned with real risk.

Minimal token overhead

Deterministic policy eval — not a second LLM. Typically under 5% token increase vs native LLM alone.

Why Agent OS

Enterprises will not adopt agents without these three things.

Auditability — every action explainable under policy and risk assessment. Risk control — high-risk work requires human approval. Accountability — signed, immutable provenance when something goes wrong.

Competitors (OpenAI Agents, Nvidia NeMo) focus on capabilities. Agent OS focuses on control. That is the enterprise gap.

Join the waitlist