HubSpot Launches Agent CLI: AI Agents Automate Marketing and Sales Workflows

HubSpot Launches Agent CLI: AI Agents Automate Marketing and Sales Workflows

In May 2026, HubSpot introduced the Agent CLI (command-line interface), bringing HubSpot's data and intelligence into the environments where go-to-market (GTM) and operations teams are composing their own workflows — Codex, Claude Cowork, and Claude Code — and letting AI agents automate repetitive, bulk, and scheduled work.

The simplest way to think about it: take the questions you keep asking, or the tasks you keep completing, in chat, and automate them. Build automations in Codex or schedule them in Cowork, and the work happens on its own before you even get to your desk.

The CLI is built on the same foundation as HubSpot's public APIs and MCP server that power its AI Connectors, and it is designed to complement them, not replace them. AI Connectors are great for work with a human in the loop: questions, insights, conversations, and campaign analytics. The Agent CLI can handle those too, but it is especially useful for the repetitive, bulk, and scheduled work that needs to run without a human in the loop.

Here is how the two divide the work:

How the Agent CLI strengthens agents working on HubSpot

The Agent CLI helps GTM and ops teams automate and schedule routine tasks, reports, and actions so they get more time back for the work that matters. For example:

• A marketer can ask for a report every Monday at 8 a.m. that surfaces high-fit contacts with no associated deal, no recent sales activity, or missing key enrichment fields, then sends RevOps a prioritized cleanup list with suggested next actions.
• Sales and RevOps can run a daily scan of the pipeline for deals closing this week with no activity in the past five days and request a summary.
• Customer Success can get an automated account review that summarizes open deals, recent support activity, and the latest NPS score for each account.
• Support can set up an automation so that whenever a new ticket arrives from a top-tier account, the agent pulls the last five tickets, summarizes each resolution, and flags recurring issue patterns.

The work happens in the background, ready when you need it.

A concrete automation example

Translated into concrete steps, it looks roughly like this: after configuring the HubSpot Agent CLI inside Claude Code or Codex, you describe the task in natural language and the agent turns it into HubSpot API calls for both reads and writes. For "every Monday at 8 a.m., find deals closing this week with no activity in the past five days," the flow could be:

1. Provide credentials: give the CLI an Access Token from a HubSpot Private App;
2. Describe the task in natural language: scan the pipeline for deals closing
   this week with no activity in the past five days;
3. The agent translates it into API calls (e.g., GET /crm/v3/objects/deals
   with the right filters);
4. Schedule it: trigger every Monday at 08:00 in Cowork and output the summary
   to your chosen channel.

The point here: you control the task description, the filter logic, and the output format, while the agent turns natural language into a repeatable workflow. Compared with logging into the CRM and filtering deal-by-deal by hand, this flow saves twenty to thirty minutes a week and never misses a key deal because someone was too busy.

Two details are worth planning ahead: first, permission boundaries — create the Access Token given to the CLI with least privilege, scoped to only the objects and operations it truly needs, rather than full CRM access; second, output validation — clean-up lists and summaries produced by automation still need spot checks by a human, especially for operations that delete or bulk-modify. Run read-only queries first, get human confirmation, and only then execute writes.

Why agent infrastructure optionality matters

HubSpot is building a platform where agent infrastructure is a choice your agents make based on what is right for your workflow. An agent constrained to a single infrastructure is less effective than it could be. Just as customers should be free to choose the best tools for their business, agents should have the same freedom — whether they are running a scheduled automation, processing a bulk operation, or acting on a real-time signal.

The direction is clear: wherever agents are working, and whatever infrastructure they run on, HubSpot supports it. The Agent CLI is available in private beta now.

Reference: HubSpot product updates and beta signup https://www.hubspot.com/product-updates; HubSpot Developer docs https://developers.hubspot.com/

16IDC perspective

For business teams, the Agent CLI answers a more fundamental question: a CRM is no longer just a system that records customer data — it can be a workflow engine driven by AI agents. When comparing CRM systems, it is worth checking whether a platform exposes open APIs, an MCP server, or a CLI, because that decides whether marketing and sales automation can truly be handed off to agents. Combined with email marketing automation, high-frequency customer operations can gradually become automated workflows. See more in the Business Tools category.

Source: https://blog.hubspot.com/marketing/introducing-the-hubspot-agent-cli