HubSpot's AI agents can do a fair bit. They'll resolve support conversations, research prospects, answer questions from your CRM data, and run workflows that used to need a person. The capability is real, and the demos show it off well.
What the demos don't dwell on is what it costs to run them - and that's the part worth understanding before you switch one on, because the pricing model changed significantly in 2026 and a lot of the advice floating around is now out of date.
So here's the honest breakdown. What HubSpot's Breeze agents actually cost, how the credit system works underneath, and the specific settings that stop a busy month turning into a bill you didn't see coming.

The Pricing at a Glance
Pricing current as of June 2026. HubSpot updates its AI pricing regularly - the last major change landed in April 2026 - so check HubSpot's own pricing pages before making a final decision, and treat the figures below as a snapshot rather than a permanent quote.
| What | Cost | How it's Charged |
|---|---|---|
| Breeze Assistant | Included, no extra cost | Free across all plans |
| Breeze Customer Agent | $0.50 per resolved conversation | Outcome-based (you pay only on resolution) |
| Breeze Prospecting Agent | $1.00 per lead recommended for outreach | Outcome-based |
| Breeze Data Agent | ~$0.10 per answer | Credits per action |
| HubSpot Credits | $10 per 1,000 (~$9 per 1,000 annual) | Consumed per action; don't roll over |
| Agent access | Generally Professional/Enterprise | Plus possible onboarding fees from ~$1,500 |
Everything below explains what these numbers actually mean in practice. And, more importantly, where the costs hide that this table doesn't show.
First, What Are We Actually Talking About?
HubSpot's AI lives under the name Breeze, and it has a few layers that get muddled together. Worth separating them, because they cost very different amounts.
Breeze Assistant is the conversational helper built into HubSpot - draft an email, summarise a record, ask a question. It's included across HubSpot plans, even free, at no extra cost.
Breeze Agents are the autonomous ones - they go off and do entire jobs on their own. The Customer Agent handles support conversations. The Prospecting Agent researches and reaches out to leads. The Data Agent answers custom questions from your CRM data. These are the ones that cost money to run, and they're generally Professional and Enterprise capabilities.
When people ask "what do HubSpot's AI agents cost," it's almost always the Agents - not the Assistant - they need to understand. So that's where we'll focus.

The Big Change: Outcome-Based Pricing
Here's the thing most existing content gets wrong, because it predates the change.
In April 2026, HubSpot moved its two flagship agents to outcome-based pricing. You no longer pay for the agent attempting the work. You pay when it completes it. That's a meaningful shift in your favour.
The Customer Agent is charged per resolved conversation - not per conversation it attempts. If it can't resolve something and escalates to a team member, you're not charged for that one. (Before the change, it was charged per conversation whether resolved or not, so the new model is both cheaper and fairer.)
The Prospecting Agent is charged per lead recommended for outreach - again, tied to an output, not just enrolment.
Both come with a free trial period, and both require a Professional or Enterprise subscription on the relevant hub.
The logic, in HubSpot's own framing: you pay when it works, not when it tries. For once, that's a pricing change that genuinely favours the customer - you're not burning money on the agent's failed attempts.
How the Credit System Works Underneath
Even with outcome-based pricing on top, there's a credit system running underneath that you need to understand - because most of the other agents still run on it directly.
Here's the model in plain terms:
Credits are bought in bulk (see the table above for the current rate), and the per-thousand cost drops slightly on an annual commitment.
Different actions consume different amounts. A resolved Customer Agent conversation consumes credits equivalent to its per-resolution price. The Data Agent costs a small amount per answer. Other agents - Content Agent, Social Media Agent, Data Agent - run on credits per action rather than outcome-based pricing.
Your account comes with included credits based on your highest subscription tier, and you can buy more when you run out.
The catch worth circling in red: credits don't roll over. Unused monthly credits expire. And all your teams draw from one shared credit pool - so support, sales, and marketing are all spending from the same bucket.
That's the part that catches people out. Not the per-action cost, which is low. The fact that consumption scales with volume, the pool is shared, and the meter is always running in the background.
Where the Real Cost Hides
If the per-action cost is so low, how do people end up surprised by the bill? Three places.
Volume scales faster than you expect. Fifty cents per resolved conversation sounds trivial. A business resolving a few hundred support conversations a month is into real money - and a high-volume support operation can spend well over a thousand dollars a month in agent credits alone before any other costs. Low per-unit cost times high volume is still a meaningful number. Model your actual volume, not the headline rate.
The platform underneath the agents. This is the bigger one. The agents that produce the impressive results generally sit behind a Professional or Enterprise subscription. If you're already on Pro or Enterprise for other reasons, adding agents is genuinely cost-effective. But upgrading your whole subscription purely to access an AI agent is rarely a smart financial decision - you're buying a large platform to use one feature. And if you want agents to work across functions - Prospecting Agent handing leads to sales, Customer Agent updating service records - you may need seats across multiple hubs.
Mandatory onboarding fees on higher tiers. Buying Professional or Enterprise directly from HubSpot typically comes with a one-time onboarding fee that can run from around $1,500 into the several-thousands depending on the hub and tier. It's one-time, but it's real, and it belongs in your first-year cost calculation.
The pattern: the per-action AI cost is low and fair. The platform and access costs around it are where the real money is. Anyone evaluating Breeze agents should be modelling the whole picture, not just the per-resolution price.
How to Avoid Bill Shock: The Settings That Matter
Here's the genuinely practical part - the specific things to do so a busy month doesn't produce a nasty surprise.
Turn off auto-upgrade. Use pay-as-you-go overage instead. This is the single most important setting. By default, HubSpot can automatically bump you to a higher credit tier when you exceed your allocation - and once you're on a higher package, you generally can't downgrade until the end of your contract. Switch to pay-as-you-go overage and you pay for what you actually use in a busy month, without getting locked into a bigger ongoing commitment because of one spike. You'll find this under Account & Billing, then Usage & Limits, then Manage credit settings.
Set a spending cap. HubSpot lets Super Admins set monthly credit spending limits. Set one. It's the hard ceiling that protects you from a runaway month.
Turn on the usage alerts. You can configure automated alerts when usage hits 75%, 85%, and 90% of your allocation. Turn all three on, especially when you first deploy an agent and don't yet know your real consumption pattern.
Watch usage weekly for the first month or two. Until you know your actual consumption rhythm, check it weekly rather than discovering it on the invoice. Once you know your pattern, monthly is fine.
Watch the beta agents. Some agents are currently free in beta, but HubSpot reserves the right to start charging with 30 days' notice. If you've built a process around a free beta agent, know that the meter could start running, and budget for the possibility.
-2.jpeg?width=400&height=264&name=_%20(5)-2.jpeg)
So, Is It Worth It?
Here's the honest answer, which depends entirely on your situation.
If you're already on HubSpot Professional or Enterprise and you've got genuine volume for an agent to work on - a real support load for the Customer Agent, a real prospecting need for the Prospecting Agent - then yes, the agents are well-priced and the outcome-based model means you're paying for results. Set your caps and alerts, and it's a strong addition.
If you're on a Free or Starter plan and thinking about upgrading purely to get an AI agent - slow down. You'd be buying a whole platform tier to access one feature, and the maths rarely works at that point. There may be a stronger case once you're on Pro or Enterprise for other reasons.
If your data lives in lots of places outside HubSpot - a separate CRM running in parallel, product data in a separate warehouse, billing somewhere else - know that Breeze reasons over data inside HubSpot. The agents are at their best when HubSpot is your genuine system of record. The more your context lives elsewhere, the more their value is capped.
The agents are good. The pricing is fairer than it was. The thing that bites people isn't the agents - it's switching them on without setting the caps, modelling the volume, or understanding what sits behind the per-action price.
Once you know what they cost, the next question is how to roll them out without breaking anything - which is its own job. We've covered that in a separate guide on deploying HubSpot's AI agents without breaking your data.
Thinking about deploying HubSpot's AI agents but want to know what they'll actually cost for your situation? Our AI solutions team models your real numbers, sets the guardrails properly, and makes sure the only surprises are good ones. Come give us a nudge.
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.
Happy HubSpotting!