One agent, for one team trying it out.
- β1 agent
- β5,000 runs a month
- βApprovals on Slack and mobile
- β30 days of replays
Relay runs agents for refunds, invoices and support, then asks a human before anything risky happens. Every step is logged, explained and replayable.
This is a live demo. Approve or reject when the agent asks.
A live map shows what each agent is doing right now, which tool it is using and how long it took. No black boxes.
Set limits like βrefunds over β¬200β and Relay pauses the agent, explains its reasoning and pings the right person on Slack or mobile.
See what the agent saw, what it decided and what it did at every step. Perfect for audits and for teaching agents better habits.
Spending limits, allowed tools and working hours are enforced outside the model, so a clever prompt canβt talk its way around them.
An illustrative estimate, not customer data. It assumes agents fully handle 62% of tickets at β¬32 an hour, and people approve the rest.
And 60+ more, plus any REST API.
One agent, for one team trying it out.
For ops teams running agents every day.
For regulated teams and many departments.
Prices in EUR, excluding VAT. Runs over your plan are β¬0.04 each.
Only inside the limits you set. Anything above a limit, outside allowed tools or outside working hours pauses and waits for a person.