What This Update Actually Is
The HubSpot Agent CLI is a command-line interface that gives AI agents, running inside Claude Code, Claude Cowork, or OpenAI Codex, the ability to read from and write to your HubSpot portal.
The keyword here is agents. Not a human typing a prompt. An agent that runs on a schedule, processes a batch of records, or fires when a condition is met. You describe the job once in natural language. The agent does it, repeatedly, on its own.
This is different from Breeze Copilot answering your questions in the HubSpot interface. The Agent CLI moves execution outside of HubSpot's UI entirely. The agent uses a CLI command set (accessible via hubspot --help after install) to interact with your data, preview changes with a dry-run flag, and take action when you approve it or when a trigger fires.
One important caveat: some actions can delete or overwrite data. HubSpot explicitly recommends using the --dry-run flag to preview changes before applying them. That's not a suggestion to ignore.
Why HubSpot Shipped This
Here's the pattern we see constantly across HubSpot portals: there's a list of recurring tasks that everyone knows needs to happen, but a human still has to initiate each one. Pipeline hygiene checks. Contact enrichment gap reports. Account summaries before QBRs. These aren't hard tasks. They're just repetitive, and they drain real time from the humans who should be using that time on decisions, not data prep.
The internal frustration HubSpot is solving: you've already learned to ask AI good questions. But you're still doing it manually, every morning, every Monday, before every review. That's not AI doing work for you. That's AI waiting for you to show up.
HubSpot's stated vision is an open ecosystem where agents can both run on HubSpot and run HubSpot. This CLI is the second half of that: agents that take action inside your portal without a human in the loop for every step.
How to Use It Step by Step
Note: Super Admins must request access to the private beta before these steps will work. Once approved, here's the setup:
- Open Claude Code, Claude Cowork, or OpenAI Codex and paste the install prompt HubSpot provides. On macOS, Linux, WSL, or Bash, it runs a curl command. On Windows PowerShell, it runs a separate PowerShell install script. Both pull from HubSpot's official API endpoint.
- Authenticate by running hubspot auth login and signing into your HubSpot account when prompted. Authorize the CLI to access your portal.
- Run hubspot --help to see available commands. Ask the agent what it can do with the hubspot command before you build your first automation.
- Write your first prompt in natural language. Use one of HubSpot's sample prompts as a template: a scheduled pipeline check, a Monday morning enrichment gap report, or a pre-call account summary. Keep the scope narrow on your first run.
- Before running any write or delete operation, ask the agent to preview using the --dry-run flag. Review the output. Confirm it matches your intent before applying changes to live data.
- Once you're confident in the output, set a schedule (daily, weekly, or event-driven) inside your agent workspace so the task runs without manual triggering.
What It Touches in Your HubSpot Strategy
The Agent CLI doesn't live inside a single hub. Its reach depends entirely on what data and actions you give the agent access to. Here's where the ripple effects land.
CRM Data Quality: Scheduled enrichment gap reports and cleanup lists hit directly at your contact and company record health. If your team has been manually auditing for missing properties on a cadence, this replaces that entirely. The agent surfaces the gaps and suggests actions. Humans approve and execute.
Sales Pipeline Hygiene: Deal health checks running at 7am every morning change how reps start their day. Instead of logging in and manually reviewing the pipeline, a rep receives a filtered summary of at-risk deals with no recent activity. This pairs well with how conditional property logic can shape deal records, which is covered in our look at
Sales Pipeline Hygiene: Deal health checks running at 7am every morning change how reps start their day. Instead of logging in and manually reviewing the pipeline, a rep receives a filtered summary of at-risk deals with no recent activity. This pairs well with how date and datetime properties now trigger conditional property logic inside HubSpot records, giving agents even more precise data signals to act on.
RevOps Processes: The Monday morning enrichment list use case is a RevOps workflow that most teams handle with manual exports, spreadsheets, and Slack messages. The Agent CLI collapses that into a single scheduled prompt. If you're also managing CRM record views for your team, recent work on the
RevOps Processes: The Monday morning enrichment list use case is a RevOps workflow that most teams handle with manual exports, spreadsheets, and Slack messages. The Agent CLI collapses that into a single scheduled prompt. If you're also managing CRM record views for your team, the streamlined CRM index page and board view update makes the data the agent surfaces even easier for reps to act on when they do log in.
Customer Success: Pre-meeting account summaries that pull open deals, support tickets, and NPS scores are something CS teams currently assemble manually before every review. That's 15 to 30 minutes per account. An agent can draft that summary automatically. The humans use that time to prepare strategy, not gather data.
Service and Support: Pulling ticket history, summarizing resolutions, and flagging recurring patterns across a contact's cases is exactly the kind of context a support rep needs before a call. The agent can generate that in seconds from your live ticket data. No copy-paste from five different screens.
Key Takeaway
The Agent CLI's value compounds over time. The first automation saves you 20 minutes a week. The fifth automation changes how your whole team operates. Start with one high-frequency, low-risk task. Build from there.
Key Takeaway
Always use --dry-run before any write or delete operation. The agent can overwrite live data. Preview first, apply second. This is not optional for production portals.
Who Should Care Most
Not everyone in your organization needs to touch this. But a few specific humans should prioritize getting into the private beta now.
- RevOps leads who currently own manual data hygiene processes: enrichment audits, duplicate reviews, pipeline health reports. This removes the human trigger from most of those workflows.
- Sales managers and sales ops professionals who want deal health visibility without building a dozen saved views or custom reports. A scheduled morning summary beats logging in and scanning manually.
- Marketing ops teams running contact database programs: identifying missing fields, surfacing high-fit contacts with no deal, or flagging segments that need re-engagement. Scheduled agent tasks replace the weekly manual export routine.
- Customer success managers at companies with large account books. The pre-review account summary use case alone justifies getting into beta early.
- HubSpot Super Admins at companies already using Claude Code, Claude Cowork, or OpenAI Codex. If those tools are already in your stack, the barrier to entry is low. The CLI install is a single command.
If your company doesn't yet use any of the three supported AI coding environments, this update isn't actionable today. That's a real constraint worth naming. You'd need to adopt one of those platforms first before the CLI has anywhere to run.
George's Take
I've been watching HubSpot portals accumulate uncompleted, recurring tasks for years. Not because the teams didn't care. Because there wasn't enough time, and the tools required a human to initiate every single step. The Agent CLI changes that architecture in a fundamental way. We're moving from AI that answers questions to AI that does the work on a schedule. That shift matters. The humans in your organization shouldn't be spending Monday mornings assembling cleanup lists or pipeline summaries from scratch. They should be walking in and making decisions from reports the agent already built. That's what flourish looks like for a modern revenue team.
“We're not talking about AI that helps you work faster. We're talking about AI that works while you're not working. That's a different category of value, and HubSpot just opened the door to it.”
Ready to Put This to Work in Your Portal?
The Agent CLI is one of several AI-powered automation tools HubSpot is rolling out in 2026. If you're also evaluating how AI can update CRM records after sales calls, our breakdown of Notetaker with Smart Deal Progression shows how these tools are designed to work together.
Getting the most out of the Agent CLI means knowing which tasks in your current operations are the right fit for automation, which data is clean enough to act on, and where human review still needs to stay in the loop. That's exactly what the Sidekick team helps HubSpot-powered organizations figure out. If you want a strategic review of where AI automation fits in your portal, book a strategy call and let's map it out together.
Frequently Asked Questions
What is the HubSpot Agent CLI?
The HubSpot Agent CLI is a command-line interface that connects AI agents running in Claude Code, Claude Cowork, or OpenAI Codex to your HubSpot portal. It lets those agents read from and write to your CRM data, run scheduled automations, and process bulk tasks without a human triggering each action manually.
Who can use the HubSpot Agent CLI?
It's available to all HubSpot customers across all paid tiers who also have access to Claude Code, Claude Cowork, or OpenAI Codex. However, it's currently in private beta. Super Admins must request access through HubSpot's product updates page before they can install or use it.
Is the HubSpot Agent CLI safe to use on live data?
Some Agent CLI commands can delete or overwrite live HubSpot data. HubSpot recommends using the --dry-run flag to preview any changes before applying them. Always test scheduled automations with a dry run first. This is especially important for bulk operations that touch contact, deal, or company records.
What kinds of tasks can the HubSpot Agent CLI automate?
Common use cases include scheduled pipeline health checks for sales reps, Monday morning enrichment gap reports for RevOps, pre-meeting account summaries for customer success, and recurring ticket history summaries for support teams. Any repeating task you currently initiate manually in HubSpot is a candidate.
Does the HubSpot Agent CLI replace HubSpot workflows?
No. HubSpot's native workflows handle record-based automation triggered by property changes or events inside the portal. The Agent CLI handles more complex, language-driven, scheduled tasks that span multiple objects or require summarization and reasoning. The two tools are complementary, not competing.
What AI tools does the HubSpot Agent CLI work with?
As of the May 2026 launch, the Agent CLI works with three AI environments: Claude Code, Claude Cowork (both from Anthropic), and OpenAI Codex. HubSpot has not announced support for other AI coding environments yet. If you don't currently use any of these three platforms, you'd need to adopt one first.
