Four guided tours — pick the question you came here with

five screens each · steps live in the URL, send a colleague the one that convinced you

CISO

Can I trust what my agents did?

You are being asked to sign off on AI agents touching production. Here is what you can actually verify.

Start the tour

Head of AI Platform

Can I run this without a new team?

You have agents to ship and no headcount for identity infrastructure. Here is what you operate, and what you do not.

Start the tour

Compliance / Risk

Can I hand this to an auditor?

Your auditor is going to ask what the agents did. Here is what you hand them, and why it holds up.

Start the tour

External auditor

Can I verify this without trusting anyone?

You were handed a package by a client. Here is how you check it without installing anything, and what you can and cannot conclude from it.

Start the tour

Or watch — nine short videos, every frame this console

All videos →

Clusters

2 healthy

Agent processes — Healthtech tenant (AWS)

all identified

Evidence rows (24h) — EKS

2,744

1,560 tool calls

Signed (24h)

2,737

100% attributable to a process

Clusters

GET /v1/clusters · GET /v1/heartbeat/history

Health	Cluster	Tenant	Cloud	Identity plane	Agents	Evidence 24h	Signed	Last beat
healthy	hexr-runtime-aks-1	globex-azure	aks	3 srv · 4 agt nested	1	4,898	3,672	37s ago
healthy	hexr-runtime-eks-1	acme-aws	eks	3 srv · 6 agt nested	2	2,803	2,803	56s ago

What this fleet is

Two runtime clusters — Azure (AKS) and AWS (EKS) — each with its own per-tenant trust root chained to the Hexr control plane. Agents on each get a process-level identity, and every action they take is a signed row in that cluster's own evidence store.

The Azure tenant runs a three-process hiring pipeline: two registered, one not. The AWS tenant runs a healthtech denial-review crew: four identified processes, one of which calls the others over agent-to-agent mTLS every three minutes.

How to read this demo

Fleet and heartbeat are read from the control plane as you look (the top bar says when). Evidence rows come from inside each cluster and are a dated capture. Every panel shows the request that produced its numbers.

Nothing here is representative data. When a cluster is unhealthy, it shows here, with the reason. Pick a tour above, or use the sidebar.