A few weeks ago, I wrote about our vision for the agent era: agents should be able to work on HubSpot, and use HubSpot. I want to dive deeper into what “run HubSpot” actually means, and our latest step to bring this vision to life.
Businesses don’t just send employees to HubSpot to get work done. They send agents. And those agents need to be able to serve you as effectively as possible, wherever they work.
That last part is important. An agent does not always work in the same environment, in the same infrastructure.
With AI connectors, HubSpot context and actions are already available in Claude, ChatGPT, and other places where teams work. Now, we add another agent infrastructure: Command Line Interface (CLI).
Introducing the HubSpot Agent CLI
HubSpot Agent CLI brings HubSpot’s data and intelligence to the places where GTM and ops teams create their workflows – Codex, Claude Cowork, and Claude Code – and allows agents to perform repetitive, bulk, and scheduled work.
A simple way to think about it: take the questions you’ve been asking or the tasks you’ve been completing over and over again in a conversation, and automate them. Build automation in Codex or schedule it in Cowork, and the work is done for you before you even get to your desk.
It’s built on the same foundation as our public APIs and the MCP server that already powers our AI Connectors — and is designed to complement them, not replace them. AI connectors are ideal for work where a human is present, talking to an agent: questions, details, conversations, campaign statistics. The Agent CLI can help agents accomplish those tasks, too, but it’s most useful for repetitive, high-volume, and scheduled work that needs to work without a human in the loop.
How the Agent CLI powers agents running on HubSpot
The HubSpot Agent CLI will help GTM and ops teams automate and schedule routine tasks, reports, and actions to get more turnaround time to do important work. No more asking for the same thing multiple times. For example:
- A marketer might request a report every Monday at 8 a.m. that brings in top-level contacts with no related deals, no recent sales activity, or key enrichment fields, and then send RevOps a prioritized cleaning list with suggested next steps.
- Sales and RevOps can have a daily pipeline scanner of deals that closed this week that didn’t close in the last five days, and ask for a summary.
- Customer Success can receive an automated account update that summarizes open deals, recent support activity, and the final NPS score for each account in the book of business.
- Support can set up an automation when a new ticket comes in from a top-level account, have the agent pull the last five tickets from the contact, summarize each fix, and flag patterns of ongoing problems.
The work happens in the background, ready when you need it.
Why it is important to choose an agent infrastructure
We’re building a platform where the agent infrastructure is the choice your agents make based on what’s best for your workflow.
An agent tied to a single infrastructure is less efficient than it could be. Just as customers should have the freedom to choose the best tools for their business, agents should have the same. Choice allows agents to choose the best infrastructure for the task at hand to be more efficient, whether they are using scheduled automation, multitasking, or real-time signaling.
The direction is clear: wherever agents work, and whatever infrastructure they work on, HubSpot supports it. That’s what an agent’s term structure looks like.
The HubSpot Agent CLI is available in private beta now, and anyone interested can sign up here.
