Setup · Web platform (browser)

For Day 2 afternoon · runs in any modern browser

Prerequisites

• A modern browser — Chrome, Edge, Safari, or Firefox (latest version)
• A printed login card (distributed at session start)
• Outbound HTTPS to the hosted platform URL

That is all. Nothing to install.

Access

1. Open the platform URL on the printed card
2. Sign in with the user name and password on the card
3. The AI Guardrail Lab company is pre-loaded with nine agents and sample incidents

Verify

Once signed in, you should see:

• Top-left company switcher showing one or more companies
• Left rail with Agents, Issues, Runs, Library
• A green status badge on each agent — the runtime is healthy

If the agent list is empty, the company switcher may be on the wrong company. Switch to AI Guardrail Lab.

Run a first task

1. Open Issues
2. Click New issue
3. Paste a one-line task — “Summarise the most recent AI incident in the library and propose one guardrail”
4. Assign to the Incident Collector agent
5. Watch the agent pick it up — comments, status changes, and a final result appear in the issue

You should see the issue move through assigned → in-progress → done within ~30 – 60 seconds.

Troubleshooting

Login fails on first attempt — confirm the case of the user name. Login codes are case-sensitive.

Pages load but agents show no status — the agent runtime is reaching out to Vertex AI. If your network blocks the endpoint, agents will hang. Switch to a network with open outbound HTTPS or fall back to the demo workspace pre-loaded at session start.

Can’t see the company switcher — your account may be scoped to a single company by default. Sign out and sign back in; or raise it at the session.

Twelve logins, more than twelve people — pair up. Two operators per login works well for the morning of Day 2.