Help centre
Agent Builder
Build agents that start on their own and work in your systems.
Getting started
Open Agent BuilderFind Agents in the menu, check you have access, and find your way around the list of agents. Then open the builder to create a new agent or change one you already have.- Build your first agent from a messageTell the AI Agent Builder what the agent should do, and it lays out and connects the steps on the canvas for you. This guide covers opening the panel, writing a request it can act on, answering its questions, checking changes before they land, refining the result, attaching files, saving and connecting accounts, and what it costs.
- Build and edit agents from chat and WhatsAppAsk Centrus to create a new agent, or change one you already have, in an ordinary chat on the web, on WhatsApp or in Teams. This guide covers the organisation switch, who can build, checking the plan, saving a draft or publishing, connecting accounts and finding the result in Agent Builder.
- Find your way round the canvasThe canvas is where you lay out an agent's steps, connect them and set each one up. This guide walks through every part of the builder, from the toolbar to the keyboard shortcuts, using one example agent throughout.
How an agent starts
Choose how an agent startsEvery agent starts from one trigger: a schedule, an email, a chat request, an alert from another system or a phone call. This guide explains what each one is for, what it costs while it waits and who has to set it up, so you can pick the right one before you build.- Run an agent on a scheduleA Schedule Trigger starts an agent at set times: every hour, at 07:00 each weekday, on the last day of the month or on dates you list. This guide covers every setting, when runs actually start, what a schedule costs, and how to test, change or pause one.
- Start an agent when an email arrivesRun an agent each time a matching email lands in an Outlook folder or Gmail label: supplier invoices into Xero, job requests into BigChange, purchase orders into a spreadsheet. Set up the mailbox and filters, see what each run receives and what it costs, and test it before it goes live.
- Start an agent from a system alertBigChange, Simpro, Joblogic, Fixflo, Freshdesk and Ostara can each tell Centrus the moment something changes, such as a finished job, an accepted quote, a new tenant repair or a new work order from a client. This guide shows how to set up an alert trigger, what the agent receives, how to test it and what it costs, then covers what is different for each system.
- Run an agent from chat and WhatsAppGive an agent a Chat Trigger and anyone with access can run it by asking for it on the Chat page, on WhatsApp or in Microsoft Teams. This guide covers setting up the trigger, what people see while the agent runs, and how the result comes back.
What an agent can do
Steps an agent can useA tour of every step in the Nodes panel apart from triggers: what each one is for, the settings that matter, how data moves between them and what each one costs. Knowledge, AI, file, data and integration steps are covered in full here; control, text and phone steps have guides of their own.- Conditions, loops and approvalsThe Control steps decide where each item goes next: down one branch or another, round a loop once per item, or into a wait while a person approves. This guide covers all ten, with their settings, what they cost and how to check which branch ran.
- Send text messagesThe Send SMS step texts a UK mobile from your organisation's Centrus number: an engineer on the way, a reminder the day before, an urgent job for the on-call manager. This guide covers every setting, the number formats it accepts, texting a whole list, what each text costs, how to test without texting anyone, and what to do when a text is not sent.
- Make outbound calls and answer incoming callsAn agent can ring people for you and answer a phone line. In both cases a Voice AI does the talking: it confirms a visit, takes a job update from an engineer, looks things up while it talks, and hands back what was said. This guide covers both directions, every setting, testing, what calls cost and what to do when a call goes wrong.
Testing and running
Test an agent before it goes liveCheck an agent against your real systems before it runs on its own. This guide covers the Spike Test step, Validate, full and dry runs from Run now, how to test each kind of trigger, and what testing costs.- Publish, pause and restore versionsAn agent only runs by itself once it is published. This guide covers publishing, changing an agent that is already live, pausing it by switching it back to draft, and going back to an earlier version from its version history.
- Read the Runs panelEvery run of an agent, step by step. This guide covers what each status means, how to read a step and its output, approving and stopping runs, the messages a failed run shows and what to do about each, and who can see runs.
Access and costs
Share an agentA new agent is yours alone until you share it. Give everyone, a team or one person View, Use or Edit access, see what each level lets them do, and check what stays with you when you share.- Automation credits and usage limitsAgents run on automation credits, a monthly allowance shared by your whole organisation. This guide covers what each step, text, call and browser session costs, when the allowance resets, the per-run budget and other limits, the warnings you will see, and where to check what each agent has used.