HubSpot Agent Hub and Agent Builder Guide: What They Do, What They Cost in AUD and Where a Person Stays in the Loop
Agent Hub is HubSpot's one console for every AI agent in your account, and Agent Builder is where you make your own. What each does, what they cost in AUD, which agents to start with, and where a person stays in the loop before an agent acts on a live customer.
HubSpot Agent Hub is one console for every AI agent in your account: HubSpot's own, ones from the Marketplace and ones you build. Agent Builder is where you build them, in plain language. Both come with Starter, Professional and Enterprise plans, and agents are paid for in HubSpot Credits, about 1.5 Australian cents each.
HubSpot launched Agent Hub and Agent Builder in public beta on 23 July 2026, for Professional and Enterprise customers at first. HubSpot's current guides, updated in October 2026, list Starter plans too. If you've been running a Data Agent inside your workflows here and a Prospecting Agent there, wondering whether either one knows the other exists, that is the gap this closes.
From HubSpot's guides to Agent Hub and Agent Builder, updated 1 and 13 October 2026.
What problem does Agent Hub solve?
HubSpot has been shipping individual AI agents for a couple of years now: Customer Agent for support, Prospecting Agent for sales outreach, various others along the way. Each one is genuinely capable. The catch is that they've been operating in their own little worlds.
HubSpot's own example of the problem is a good one: a Prospecting Agent reaches out to a contact the same week a Customer Agent is handling an open complaint from that same account, and neither agent has the faintest idea the other is talking to that customer. One arm doesn't know what the other is doing, which is exactly the kind of failure a team of people would never let happen, and these tools are supposed to be smarter than that.
Agent Hub is the fix. A single console where every agent running across your go-to-market shares the same customer context, instead of each one working from its own narrow slice of the picture.
What is HubSpot Agent Hub?
Agent Hub is best understood as the new home base for AI agents in HubSpot, in fact, it's the evolution of what used to just be called "Breeze Agents," now given its own proper hub rather than being scattered across other product areas.
Here is what it gives you in one console, according to HubSpot's Agent Hub guide:
See what's running. Every agent active in your account: the pre-built ones from HubSpot, anything you've installed from the Marketplace, and anything custom you've built, organised by the outcome it's driving. HubSpot's own framing groups agents by goals like building demand, winning deals, delighting customers, and growing the account, where they used to sit under whichever hub they belonged to.
Check performance. Instead of digging through separate reporting for each individual agent, Agent Hub gives you visibility into what's actually happening, what's being resolved, what's being escalated, what's working.
Turn things on. If HubSpot has agents available on your plan that you haven't activated, they're visible here rather than buried in a settings menu somewhere.
Reach the Marketplace without leaving. Pre-built agents you might want to install are accessible directly from Agent Hub, with no separate trip through HubSpot's wider app marketplace.
Give agents your business context. A Context tab holds business context and knowledge vaults that agents draw on, so each one is not working from its own notes.
The genuinely useful shift is that "which agents do we have, and are they doing anything sensible" finally has one honest answer instead of five scattered ones.
What is HubSpot Agent Builder?
Agent Builder is the other half of this release, and it's the part with real hands-on substance. It's also had a rename: this is what used to be called Breeze Studio.
Agent Builder is where you build and improve your own custom agents and automations, on a single canvas, in plain language, with no coding required. You describe what you want the agent to do, and it builds itself from that description, working from data that's already sitting in your CRM: contacts, deals, conversations, call transcripts, whatever context is relevant.
A few things worth knowing about how it actually behaves:
Agents and automations sit on the same canvas. You're not switching between a workflow tool and a separate AI tool, triggers, actions, and agent hand-offs connect on the same canvas. Work that used to take a dozen manual steps can genuinely become one connected build. If you're not sure whether a job needs an agent at all rather than a plain workflow, that's the actual first question, and we cover it in HubSpot AI agent or simple workflow.
Agents start from workflows, automations or CRM cards. On Professional and Enterprise, an agent runs as a step in an Agent Hub workflow, so any workflow trigger can start it: a schedule, a property change, a form or a webhook. Starter accounts run agents from automations in the side panel. You can also put an agent on a CRM card so a rep runs it from the record.
You can test before you pay, and cap what it spends. Test runs in the builder do not use credits. Each live run shows its estimated credits in the run history, and you can set a monthly run limit.
A person can review what it produces. Agent output can be reviewed in the agent inbox. Use that review step while you are building trust in what the agent does, and only loosen it once the run history shows it behaves the way you expect.
It reuses your existing playbooks, in the sense that agents are built on the context already in your CRM rather than starting from a blank slate every time, the same records, history, and data your team already relies on.
One useful distinction while we're here: HubSpot separates agents from assistants. Assistants (accessed through Breeze Assistant) answer questions and provide guidance quickly, think of them as the fast, conversational helper. Agents, built and customised in Agent Builder, actually complete structured tasks and take actions in your account, which is slower because it's doing more. Worth knowing which one you actually need for a given job.
Can you install ready-made agents instead of building them?
Not sure which agents are already running in your portal?
Most were switched on by different people, months apart.
Agent Builder isn't only for building agents from a blank canvas. HubSpot's Marketplace includes pre-built agents you can install directly, then customise inside Agent Builder to fit your own business rather than starting from nothing.
Worth knowing before you install one: HubSpot-built agents come with default inputs, instructions, and tools that can't be edited, deleted, or rearranged. Agent Hub does let you clone an agent, which is the place to start if you need to change more. You're customising around the edges of a pre-built agent. For a lot of common jobs, that's a genuinely sensible starting point. Install, adjust what's adjustable, and see how it performs before deciding whether you need something fully custom instead.
This matters for how you approach the whole thing. The instinct with any new AI capability is to imagine the most elaborate possible use case and try to build it immediately. A better first move is usually the boring one: check what's already built, install it, and see whether it does eighty percent of the job before you invest time building the remaining twenty percent from scratch.
What do Agent Hub and Agent Builder cost in AUD?
The console and the builder come with your plan. The agents' work is what costs money, and it is paid for in HubSpot Credits, which we price in AUD in a separate post. Past the credits included in your plan, each credit is US$0.010, about 1.5 Australian cents, and Professional plans include 3,000 credits a month.
| Agent | What counts as one use | Credits | About, in AUD |
|---|---|---|---|
| Customer Agent | One resolved conversation | 50 | A$0.73 |
| Prospecting Agent | Recommended outreach for one lead | 100 | A$1.45 |
| Data Agent | A response to one prompt for one record | 10 | A$0.15 |
| Custom agent from Agent Builder | One run | Varies with the job | Check the run history |
AUD figures convert HubSpot's US rate at the Reserve Bank rate on 9 November 2026, before GST, and Australian accounts are usually invoiced in AUD. Customer Agent and Prospecting Agent moved to outcome-based pricing in April 2026, so you pay for a resolved conversation or a recommended lead.
Custom agents are the open question. HubSpot says a custom agent uses credits when it does work in a run, such as reasoning, taking actions or generating content, and that the cost varies with the complexity of the job. So test it, read the estimated credits per run, multiply by the runs you expect in a month, then set the monthly run limit to match.
Which HubSpot agents should you start with?
Start with the agent whose mistakes are cheapest to catch. In practice that order usually looks like this:
- Data Agent first. It fills or summarises fields on records you already own, at about 15 Australian cents an answer, and a wrong answer sits in a property where someone can check it before it reaches a customer.
- Customer Agent next, on a narrow inbox. Point it at the questions your team answers the same way every week, with a clean knowledge base behind it. Our AI agent cost post has the credit maths and our setup video.
- Prospecting Agent last. It writes to people who have not bought from you, so the Spam Act and your brand are on the line. Read whether an AI agent can replace an SDR in Australia before you let it send.
- Custom agents when a job is clear. If a plain workflow can do it, use the workflow. Choosing between a HubSpot AI agent and a workflow covers the test.
Where does a person still have to be in the loop?
HubSpot's own launch material is about visibility and coordination between agents, and it should be. What it spends less time on is the decision underneath both products: which agents are allowed to act on a live customer without a person checking first, and which ones should not be.
The Australian Signals Directorate's Cyber Security Centre guidance on adopting agentic AI services carefully is the current word on this, and it treats an agent's autonomy the way you'd treat any other production change: something to expand deliberately, with a person accountable for the decision, not something to set to maximum in the first week because the button exists. Reviewing agent output in the agent inbox, and capping runs, are the controls HubSpot gives you for this. They only work if someone actually sets them, rather than accepting whatever the default is when the first custom agent ships.
There's a second reason to get this right that isn't optional. From 10 December 2026, the Australian Privacy Principles pick up a transparency obligation for automated decision making. An agent that resolves a support conversation, or recommends a lead and moves it forward, is making a decision about a person. If that decision significantly affects someone's rights or interests, it belongs in your privacy policy, and knowing plainly which agents can act without a person and which can't is the starting point for answering that honestly.
Our own rule, covered in how to deploy HubSpot AI agents without breaking your data and how we test an AI agent before it goes near a customer, is that an agent proposes and a person approves, until the evidence says otherwise. Multiply that across every agent sitting in Agent Hub, and "have we coordinated our agents" turns out to be the smaller question. The bigger one is whether you can say, for each agent, whether a person is still in the loop.
What should you do first?
If you're already running HubSpot agents, here's the useful next step: go and actually look at Agent Hub. Most businesses running Customer Agent or Prospecting Agent have never had one dashboard showing what those agents are doing side by side, this is your first real chance to see the whole picture rather than piecing it together from separate reports.
If you've been meaning to build something custom but didn't have the technical resource to justify it, Agent Builder's natural-language, no-code approach genuinely changes that calculation. Describing a task in plain English and having it build the automation is a much lower bar than it used to be, worth revisiting anything you'd previously dismissed as "we'd need a developer for that."
And if you're running multiple agents that have never been coordinated, which, per HubSpot's own framing, is a genuinely common problem, this is the moment to check whether your Prospecting Agent and your Customer Agent are stepping on each other's toes with the same accounts. That coordination gap is exactly what this release exists to close, but only if someone actually goes and looks.
HubSpot has changed both products more than once since July, adding Starter plans, a Context tab and run limits. Treat what you build now as something you will revisit, and check the run history every month.
Frequently asked questions
What is HubSpot Agent Hub? Agent Hub is one workspace in HubSpot for viewing, switching on, configuring and monitoring every AI agent and agentic workflow in your account: HubSpot's own agents, ones installed from the Marketplace and custom agents you build.
Which HubSpot plans include Agent Hub? HubSpot's guide lists Starter, Professional and Enterprise editions of Marketing, Sales, Service, Data and Content Hub, plus Professional and Enterprise Smart CRM and Revenue Hub. HubSpot Credits are required, and some agents and agentic workflows need Professional or Enterprise.
What does a HubSpot AI agent cost in AUD? Agents use HubSpot Credits. Past your included credits, each credit is US$0.010, about 1.5 Australian cents. A resolved Customer Agent conversation uses 50 credits, about A$0.73 before GST, and Professional plans include 3,000 credits a month.
Do custom agents built in Agent Builder use credits? Yes. A custom agent uses credits when it does work in a run, such as reasoning, taking actions or writing content, and the cost varies with the job. Test runs in the builder do not use credits, and you can set a monthly run limit.
Is Agent Builder the same as Breeze Studio? Yes. Agent Builder is the new name for what HubSpot called Breeze Studio, and Agent Hub grew out of what used to be called Breeze Agents.
Wrapping up
Agent Hub and Agent Builder are HubSpot finally doing the organisational housekeeping its own agent sprawl needed. One place to see what's running, one canvas to build what's next, and a cost model that charges for completed work rather than mere existence.
If you're running HubSpot AI agents and you can't currently answer "what are they all doing, do they know about each other, and which of them can act without a person", that's the gap this release was built for.
Settle the Privacy Act question before you switch one on
An agent that reads and writes CRM records is using and disclosing personal information, which is what APP 6 of the Australian Privacy Principles covers. Decide which objects it may read, which it may write and who sees the output, then write that down before the build rather than after somebody asks.
Weighing up building an agent against buying one?
The credit cost usually decides it before the features do.
Want a hand making sense of Agent Hub and building something genuinely useful in Agent Builder, with the oversight built in from the start? Our AI engineering team can help, and we'll build the ones you don't have yet. Let's chat.
Follow our Facebook page for practical HubSpot tips, quick wins, and fixes you can use immediately. Subscribe to our YouTube for step-by-step HubSpot tutorials and real examples from the field.