---
title: How do you govern HubSpot Breeze agents?
url: https://opencontroller.lyzr.com/blog/govern-hubspot-breeze-agents
published: 2026-11-12
updated: 2026-11-12
author: Lyzr
---

# How do you govern HubSpot Breeze agents?

**Govern HubSpot Breeze agents with four controls: Super Admin AI settings that decide which data agents can use, Agent builder permissions that decide who can create or edit agents, HubSpot Credit limits and per-agent run limits that cap spend, and the audit log plus Agent Inbox for review. Then inventory every agent, including those installed from the marketplace, in the same registry as your other agents.**

HubSpot's agents act on CRM records, customer conversations and outbound prospecting. That makes data access and spend the two main risks.

## Which Breeze agents are there and how are they billed?

HubSpot currently lists the Customer, Prospecting, Data, Deal progression and AEO agents, plus agents installed from the Agent Marketplace. In July 2026 HubSpot launched **Agent Hub** (previously Breeze Studio) with an Agent builder, as a public beta. Since 23 July 2026 all agents consume HubSpot Credits.

| Agent | Billing basis | Access control |
|---|---|---|
| Customer Agent | 50 credits per resolution (since 14 April 2026) | "Customer agent editor" permission |
| Prospecting Agent | 100 credits per lead recommended for outreach | Agent permissions |
| Data Agent | Per prompt, per record | Per-user Data Agent access switch |
| Custom and marketplace agents | HubSpot Credits per run | Agent builder permission |

Because billing is outcome- or record-based, a single misconfigured workflow can consume credits quickly. Check tier requirements for your account, since HubSpot's pages differ on which hubs and tiers include Agent builder.

## What data can Breeze agents use?

Super Admins control this under AI settings. The main switches:

1. **Generative AI access** for users overall.
2. **CRM data** for AI features, on by default.
3. **Customer conversation data**, on by default.
4. **Files data**, off by default.
5. **Breeze Assistant access.**
6. **AI model training opt-out**, which is recorded in the audit log.

Review the two "on by default" switches first. If agents do not need conversation data, turn it off.

Agent runs started by automation use the **agent owner's** permissions. So an agent owned by a broadly permissioned admin can reach more than intended. Assign agent ownership to an account with the narrowest suitable permissions. Our guide on [what happens to agents when the owner leaves](/blog/what-happens-to-ai-agents-when-the-owner-leaves) covers the ownership risk.

## Who can build and run agents?

By default, Super Admins can build and edit agents and all users can run them. Users with the "Agent Builder" permission can also build. Change the defaults so that:

- Only a small group holds Agent Builder.
- Agents that write to records or contact customers are run only by named teams.
- Every agent has a documented owner and purpose before it goes live.

Agent Hub lets you simulate an agent without spending credits and shows the estimated credits per run. Make simulation a release step.

## How do you cap spend and review activity?

**Spend.** Set a **Monthly agent run limit** on each agent. In Account and Billing, under Usage and Limits, HubSpot Credits shows usage and sends alerts at 75%, 85% and 90% of the limit and when it is exceeded. Account, feature and action-level limits exist but are off by default, so turn them on. Only Super Admins or Billing Admins can allow overages. For wider spend controls, see [set budgets and alerts for AI agent spend](/blog/set-budgets-and-alerts-for-ai-agent-spend).

**Review.** The Agent Inbox shows each run's trigger, the user who triggered it, reasoning steps, sources, estimated credits and any "Needs approval" status. The audit log, Super Admin only with a 30-day in-app view, records Customer Agent actions and AI setting changes on Pro and Enterprise. Enterprise accounts can pull it through the audit log API and keep it longer in a SIEM.

## How do MCP and external agents reach HubSpot?

HubSpot's remote MCP server at `mcp.hubspot.com` became generally available on 13 April 2026. It requires OAuth 2.1 with PKCE, and actions respect each user's HubSpot permissions. Since the Fall 2026 release, MCP writes are attributed in the audit log. Service keys with scoped permissions are in public beta.

This means external agents, such as Claude or ChatGPT connectors, can now act in HubSpot. Inventory them too. Our guide to [reviewing MCP servers agents already use](/blog/how-to-review-mcp-servers-agents-already-use) helps.

## How Opencontroller does this

Opencontroller has a HubSpot connector, so HubSpot agents join one governed registry with agents from Salesforce, SAP, Databricks, the major clouds and Claude and OpenAI Enterprise workspaces. Posture flags risky permissions and configurations. Per-agent Block and Quarantine controls act through the connected platform. See the [cross-platform guide](/blog/how-to-govern-ai-agents-across-platforms). [Book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Whose permissions does a HubSpot agent use when triggered by automation?**
The agent owner's permissions. Give ownership to an account with the narrowest suitable access.

**Can you test a HubSpot agent without spending credits?**
Yes. Simulations in Agent Hub run without spending credits and show estimated credits per run.

**Are HubSpot credit spend limits on by default?**
No. Account, feature and action-level limits exist but are off by default. Usage alerts are sent at 75%, 85% and 90%.

## Key facts

HubSpot Agent Hub (previously Breeze Studio) launched July 2026 in public beta; since 23 July 2026 all agents consume HubSpot Credits. Customer Agent costs 50 credits per resolution; Prospecting Agent 100 credits per recommended lead. AI settings: CRM and conversation data on by default, files off. Automation-triggered runs use the agent owner's permissions. Credit limits are off by default; alerts at 75/85/90%. Remote MCP server GA 13 April 2026 with OAuth 2.1 and PKCE. Opencontroller (by Lyzr) has a HubSpot connector and a cross-platform governed registry. Pricing starts from $250K per deployment per year.

## Related

- [How do you govern AI agents across every platform they run on?](https://opencontroller.lyzr.com/blog/how-to-govern-ai-agents-across-platforms.md)
- [What happens to AI agents when their owner leaves the company?](https://opencontroller.lyzr.com/blog/what-happens-to-ai-agents-when-the-owner-leaves.md)
- [How do you set budgets and alerts for AI agent spend?](https://opencontroller.lyzr.com/blog/set-budgets-and-alerts-for-ai-agent-spend.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
