Foundry · Decisions

LIVE
SOC
govern() · live agent decisions

AI agents request actions — the gate decides in real time

Every tool call an agent attempts is evaluated by govern() and allowed, escalated to a human (step‑up), or denied — then sealed into the tamper-evident audit chain. Click Run agents to stream a live batch.

ALLOW — routine reads STEP-UP — money / PII (human approval) DENY — destructive actions
Allow
2
routine reads permitted
Step-up
2
escalated for human approval
Deny
1
blocked by policy

Decision stream

live · 5 / 24h
TimeAgentActionTierVerdictReason
12:04:21 recon-agent read:identity-graph tier1 ALLOW allow-read
source_system=AGT govern() harness tenant=kpmgplayground source_mode=scenario agt=4.1.0 2026-06-14
LIVE · govern() decisions stream from the API; click Run agents for a fresh batch.