Hanzo World
World Monitor — global AI-system observability
A live map of every model, agent, and AI workload across your organization. Latency, cost, drift and anomalies, shown as they happen, with alerting and forensic replay.
Paste and ship
Paste this into any agent. It reads the skill manifest and calls World from there.
See Every Model, Everywhere
Every workload you run, on one map.
Live Topology
A real-time map of every model, agent, and tool call. Trace any request from edge to GPU.
Latency + Cost
P50/P95/P99 by model and route. Per-org and per-user spend. Auto-anomaly on cost spikes.
Drift Detection
Embedding drift, output drift, prompt drift. Alert when production diverges from baseline.
Eval Dashboards
Continuous evals on live traffic. Pin scores. Compare model versions in production.
Multi-Cloud
Pulls signals from Hanzo Cloud, AWS, GCP, Azure, and self-hosted clusters. One unified view.
Forensic Replay
Reconstruct any agent session step by step. Inspect tool calls, prompts, model outputs, and IO.