HubSpot Agent Hub: Who Manages Your AI Agents?

HubSpot now gives you a dashboard for your AI agents. What it doesn’t give you is someone to run them.

That gap is easy to miss when the demo looks good. This post covers what Agent Hub and Agent Builder do, what they leave to you, and how to set up ownership before the first agent goes live. 

This post is part of our HubSpot UNBOUND 2026 coverage.

What Are Agent Hub and Agent Builder?

HubSpot launched Agent Hub and Agent Builder in public beta on July 23, 2026, for Professional and Enterprise customers. Agent Hub is one place to build, monitor and manage AI agents across marketing, sales and service, all working from shared customer context. 

Agent Builder is how you create custom agents: you describe the task in plain language and HubSpot assembles it.

As of this writing, HubSpot’s own product page lists Agent Hub itself as included in every edition: Starter, Professional and Enterprise. That’s broader than the July beta. It’s the dashboard that’s now widely available; a Professional or Enterprise subscription is still required to use some of the pre-built agents you’d manage inside it, so check what your own plan unlocks before assuming everything in this post applies.

HubSpot lists the pieces like this:

• A dashboard showing live status and performance for every active agent.
• Plain-language building through Breeze Assistant.
• A canvas that connects workflows, custom agents and triggers.
• Triggers from schedules, webhooks or third-party integrations.
• A shared foundation of deal history, contact records and buying signals.

At UNBOUND, the rebuilt Breeze Assistant brought Agent Builder into the conversation. Per recap of the keynote, you describe the agent you want, review the plan Breeze proposes, approve it, and from then on you can call that agent like a teammate.

What Agent Hub Shows You (and What It Doesn’t)

What it shows: what’s running, whether it’s working, and how agents share context.

What it doesn’t decide:

• Who owns each agent.
• What “good” looks like for its output.
• How often anyone reviews it.
• When it gets retired.
• Who’s accountable when it gets something wrong.
• How much it’s allowed to spend.

That’s not a criticism. It’s a dashboard. But dashboards only help if someone looks at them.

HubSpot Credits: What Agent Usage Costs

Several HubSpot agents run on credits rather than being included with your seat. 

As one concrete example: Customer Agent is priced at $0.50 per resolved conversation on Professional and Enterprise plans, billed through HubSpot Credits. In April 2026, HubSpot moved two of its flagship agents, Customer Agent and Prospecting Agent, to this kind of outcome-based pricing, meaning you pay when the agent completes the task rather than on a flat recurring charge.

Usage is climbing. In its Q1 2026 earnings call, HubSpot’s CEO said total credits consumed grew 67% quarter over quarter, with Customer Agent, Prospecting Agent and Data Agent making up most of that consumption.

Why that matters for a small team: cost now scales with activity, not headcount. An agent that runs too often, on the wrong records or with a weak instruction costs real money and produces bad output. 

HubSpot’s own materials describe governance controls such as usage caps and alerts, though we couldn’t confirm the exact thresholds from an official source in time for this post we advice checking your own portal’s credit settings for what’s actually configurable today, and set limits before you switch on anything metered.

Who Should Own Your HubSpot Agents?

Every agent needs three roles, even if one person fills all three:

• A business owner, who decides what the agent is for and whether it’s earning its keep.
• A technical owner, who configures it, connects it to data and fixes it when it breaks.
• A reviewer, who checks the output and approves changes.

In a small B2B team, the first is usually a founder or head of sales, and the second and third sit with someone in ops or RevOps. If you don’t have that person, that’s the gap. It’s the same gap that produces the abandoned workflows in most portals. Agents just make it faster and pricier.

Monitoring, Tuning and Retiring Agents

A simple cadence works for most teams:

• First 30 days: review output weekly. Track what gets approved, edited and rejected.
• Ongoing: check monthly on cost per outcome, error patterns and whether the agent’s job still matches your process.
• Retire: when an agent’s job changes or it stops earning its keep, turn it off and write down why. Agents nobody remembers are the new orphan workflows.

Track approval rate, override rate, cost per outcome and downstream errors like duplicate records or wrong stage changes.

And remember that an agent is only as good as the context it reads. Review that too. We cover how in HubSpot Context Home: Is Your CRM Ready for AI?

When to Hand It to a Managed Services Partner

If you have a RevOps person with time, keep this in-house. You don’t need a partner to do this well. You need an owner.

If you don’t have one, or yours is already stretched, this is what managed services exists for: someone who owns the portal month to month, reviews what’s running, tunes it and reports on it. At Manyflow, Managed Services is a monthly RevOps retainer in three tiers. 

We usually start with a HubSpot Audit, so you know what state the portal is in before agents start acting on it.

Wondering who should own this on your team? Book a free consultation and we’ll talk it through.

FAQ