What This Update Actually Is
Agent Builder is HubSpot's new environment for creating fully custom AI agents. It ships with two core pieces: a custom agent builder and an agentic automation canvas.
The agent builder is where you define what your agent knows, how it behaves, and what tools it can use. The automation canvas is where you connect that agent to triggers and actions so it can run at scale across your CRM records.
A few important boundaries to name upfront. Agent Builder is available to HubSpot Pro and above. Each agent run consumes HubSpot credits. HubSpot's own featured agents (Customer Agent, Prospecting Agent, Data Agent) are separate and not affected by this release.
If you had Breeze Marketplace agents installed before July 23rd, they've been automatically converted to templated custom agents. A small number of templated agents are being sunset, but nothing you configured is lost.
Why HubSpot Shipped This
The problem wasn't that HubSpot lacked AI features. The problem was that those features weren't yours. They were generic. They didn't know your process, your language, or your data.
Humans running busy portals have been stuck choosing between helpful-but-generic AI and custom-but-complex code. Neither felt like a real solution for a growing business without a developer on staff.
Agent Builder closes that gap. You describe what you want, Breeze Assistant builds the agent structure for you, and the agentic automation canvas connects it to your existing workflow logic. The AI finally works the way your business does, not the way a product demo assumes it does.
How to Use It Step by Step
- Confirm permissions. Any user who needs to build, activate, or edit an agent must have Agent access turned on in their user settings. Agentic workflow editing requires Edit or Super Admin permissions. Publishing requires Publish permissions for workflows.
- Open Agent Builder in your portal and describe your agent in plain language. Breeze Assistant reads your intent and scaffolds the agent for you, pulling context from the data already in your CRM.
- Customize the agent's instructions, tools, and knowledge vault. This is where you give the agent your specific playbook: which properties to read, what tone to use, which knowledge base articles to reference.
- Set a run limit on the agent details page. This caps how many times the agent can run so you don't burn through credits unexpectedly. You can also monitor total spend across all agents in one dashboard.
- Use simulation mode to test your agent before it goes live. Simulation mode shows you the estimated cost per run with zero credits consumed. This is your safety net before you scale.
- Connect the agent to the agentic automation canvas. Add triggers, define what actions the agent should take on records, and set follow-up steps. If you previously had triggers on Breeze Marketplace agents, check your Workflows app. Most were migrated automatically on July 23rd in an unpublished state.
- Handle any non-migrated triggers manually. Recurring schedule triggers for agents without a specific CRM object (like a blog creation agent) were not migrated. You'll need to recreate those in the new agentic automation builder.
- Publish when you're ready. Your migrated workflow triggers sit in an unpublished state. Turn them on when you've confirmed everything looks right.
What It Touches in Your HubSpot Strategy
Agent Builder doesn't live in one hub. It touches your entire front office.
On the CRM side, agents read and act on contact, company, deal, and ticket records. That means any property, association, or lifecycle stage you've defined becomes raw material for the agent's logic.
On the automation side, the agentic canvas is layered on top of the existing workflows engine. If your team is already using workflows for lead rotation or follow-up sequences, agents can now plug directly into those flows as intelligent decision-making nodes rather than simple if-then branches.
For example, if you're already using HubSpot's updated lead workflow rotation with Round Robin and Balanced distribution, you can now layer a custom agent on top to enrich the record or send a personalized follow-up before the rep ever touches it.
Key Takeaway
The agentic automation canvas doesn't replace your existing workflows. It extends them. Think of agents as intelligent steps you can drop into any workflow you've already built.
On the knowledge and AI side, the agent's knowledge vault ties directly to what Breeze already knows. You can feed it your playbooks, your product documentation, your sales scripts, and your support content. The agent won't hallucinate a process that doesn't exist in your portal.
Credit consumption is the new variable to track in your RevOps reporting. You'll want to add agent run costs to your operational dashboard alongside workflow usage and API calls. HubSpot surfaces estimated costs at the inbox level and on CRM card views, but your team should build a habit of reviewing usage monthly.
If you want the full strategic picture of where HubSpot's AI infrastructure is heading, the HubSpot Agentic Platform pillar covers where this has been, what's available today, and George's predictions for what comes next.
Key Takeaway
Watch your credit budget. Simulation mode is not optional for new agents. Run it every time before you publish to avoid surprise charges at scale.
Who Should Care Most
Not every HubSpot Pro customer needs to build a custom agent on day one. But certain roles and company profiles should move quickly.
- RevOps leaders who already have mature workflow libraries. Adding agents to existing automation gives you more intelligence without rebuilding what works.
- Marketing ops teams running high-volume contact nurture. Agents can personalize follow-ups at a scale no human queue can match.
- Sales ops teams managing complex deal routing or post-demo follow-up. A custom agent trained on your sales playbook can handle first-touch enrichment and outreach triage automatically.
- Service teams handling ticket triage at scale. An agent that knows your knowledge base and your escalation criteria can pre-classify and respond to tickets before a human reviews them.
- Business owners who previously avoided AI tools because the setup felt too technical. Breeze Assistant's plain-language build process removes that barrier entirely.
If you're on Starter or Free, this one isn't available to you yet. Pro is the entry point.
George's Take
I've been watching HubSpot build toward this for a while now, and what strikes me most isn't the technology. It's the permission it gives humans inside growing businesses. For years, the pattern I saw across portals was the same: great data sitting unused because nobody had the bandwidth to act on it fast enough. Agent Builder changes that equation. You're not hiring a developer. You're not buying a separate AI platform. You're teaching HubSpot how your business thinks, in plain language, and then letting it run. That's a meaningful shift. And the simulation mode isn't just a nice-to-have. It's the reason smart teams will actually trust this at scale.
“You're not hiring a developer. You're teaching HubSpot how your business thinks, in plain language, and then letting it run. That's a meaningful shift.”
If you want to see how Breeze Assistant handles intent-based building before you dive into Agent Builder, check out the live Breeze Assistant prompt demos from the team that built and teaches it. It'll show you exactly how plain-language prompts translate into real CRM actions.
Ready to build your first custom agent the right way? The Sidekick team works with growing companies to design agentic workflows that actually match how your business runs, without burning through credits on guesswork. Book a strategy call and let's map it out together.
Frequently Asked Questions
What is HubSpot Agent Builder?
HubSpot Agent Builder is a no-code environment inside HubSpot where you can create custom AI agents trained on your own instructions, tools, and CRM data. It includes an agentic automation canvas that connects agents to triggers and workflow actions so they can take action on records at scale. It's available to HubSpot Pro and Enterprise customers.
How does credit usage work with custom agents in HubSpot?
Each custom agent run consumes HubSpot credits from your plan's credit allotment. You can see the estimated cost per run in your inbox and on CRM card views. You can set run limits per agent on the agent details page and monitor total spend across all agents. Simulation mode lets you test agents and estimate costs without consuming any credits.
What happened to Breeze Marketplace agents after July 23, 2026?
Breeze Marketplace agents were automatically converted to templated custom agents on July 23, 2026. No action was required. Most triggers were migrated to the Workflows app in an unpublished state. A small number of templated agents were sunset, but all configurations were preserved. Recurring schedule triggers without a specific CRM object were not migrated and need to be recreated manually.
Do I need to know how to code to use HubSpot Agent Builder?
No. Agent Builder requires no code. Breeze Assistant lets you describe what you want your agent to do in plain language, then builds the agent structure for you based on your intent and the data already in your CRM. You then customize the agent's instructions, tools, and knowledge vault through a visual interface.
Who needs permission to build and publish agents in HubSpot?
Users need Agent access turned on in their settings to activate, credit, and edit custom agents and their tools and knowledge vaults. To edit agentic workflows, users need Edit permissions or Super Admin status. To publish agentic workflows, users need Publish permissions for workflows. Permissions are managed separately for agents and workflows.
How is the agentic automation canvas different from regular HubSpot workflows?
Regular HubSpot workflows use if-then logic to automate actions based on preset conditions. The agentic automation canvas layers AI agents into that flow as intelligent decision-making nodes. Agents can read context, use knowledge, and take nuanced actions that a traditional workflow branch can't replicate. The canvas connects agents, triggers, and actions in one unified view.
